i e :3,
3%... r:

:1!

 

 

 

‘.h.\».u~.{v ¢
“11.3.! $4 , a
6.5.1.. {11,313.
.e £325.11}.
Wk E?

v!
A.»

(‘3‘. (3:31.55:
Si: “35'. igfi .
Java}. .
.. . 5... ‘vs AI‘
3. :5 11.3.2... .

Jan

u9vlns. LP;
5!.0‘.

. ‘ ....N.. 1-.. 1 h . .
.u‘ ‘L- ‘ I I 'K" \
Gig . ghummmi... .va

 

' 3 LIBRI WY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled

ANALYZING THE RELATIONSHIP BETWEEN SITE
OPERATIONAL FACTORS AND CONSTRUCTION WORK
FLOW RELIABILITY: AN SEM APPROACH

presented by

SAMARTH JAIN

has been accepted towards fulfillment
of the requirements for the

Master of degree in Construction Management

 

 

 

«(Elli/all QM

 

Major Professor’s Signature

Aw} l’tfazouz

Date

MSU is an Affinnative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 KIProj/AooaPreleIRCIDatoDuojndd

 

ANALYZING THE RELATIONSHIP BETWEEN SITE
OPERATIONAL FACTORS AND CONSTRUCTION WORK FLOW
RELIABILITY: AN SEM APPROACH

By

Samarth Jain

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Construction Management

2010

ABSTRACT

ANALYZING THE RELATIONSHIP BETWEEN SITE OPERATIONAL
FACTORS AND CONSTRUCTION WORK FLOW RELIABILITY: AN SEM
APPROACH

By
Samarth Jain

The lack of an explanatory understanding of factors giving rise to low/high PPC
highlights the importance of investigating the current production metrics implemented in
Last Planner System and how these metrics affect workflow reliability in the production
stage. The overall goal of this research is to understand causal relations related to
workflow reliability at the production level in a construction project. To approach this
goal, the research focused on developing a method to investigate the causal relationship
between production constraints and their impact on workflow reliability as measured by
the PPC metric. The research has concluded that the latent factors (Pre Requisite Work,
Directives, Burden etc.) in construction management research are mostly subject to
constructivist interpretation, i.e., they form as a result of collection of a set of measured
variables and represent a collective existence of those variables, therefore, it is
recommended that future researchers consider it strongly and test relationships with
formative latent variables. It was also found that studying the impact of all factors
together is more insightful than isolation studies. This research has developed a
framework using which industry professionals can measure the impact of production
delay factors on work flow reliability. This research contributes to the lean production
management practices by developing a list of production factors that can be added to the
constraint analysis sheet of the Last Planner and used to record the status of production

on a quantitative scale, on a regular basis.

Dedicated to
My Grandfather, Late Mr. Padam Singh Jain,
My parents, Mr. Satish Chander Jain and Mrs. Usha Jain,
My aunt and uncle, Mrs Nirmal Jain and Mr. Ari K. Jain.

 

iii

ACKNOWLEDGEMENTS
Finally! To be able to write this page gives me utmost joy as it marks the end of this
painful but enriching experience of writing a master’s thesis. It also marks the end of my
education and, sadly, active association with Michigan State University. I would like to
use this space to express my thanks to all the people who were directly or indirectly
involved in making my journey at MSU and writing this thesis, a fruitful experience. And

I will do this without worrying about adding citations and footnotes.

To my Uncle and Aunt, Mrs. Nirmal Jain and Mr. Ari K. Jain, I thank them for putting
immense faith in me and my capabilities as a student and instilling the belief in me that I
can come to MSU against all odds and fulfill my dreams, without their support this

journey would not have been possible.

To my Advisor, Dr. Tariq S. Abdelhamid, no form of writing will be able to truly express
my gratitude and appreciation for him. He was always there when I needed a professional
advice or just someone to talk to. His enriched and simple thought process always helped
in putting me back on the tracks when I derailed from my mission. His expert guidance
and unfettering support made this rollercoaster of a journey look easier. His mentoring
during my entire master’s education and thesis writing, taught me a lot of valuable
lessons of life, which I would forever strive to follow to become a better person. One of
my greatest personal enrichments at MSU is my association with Dr. Tariq which I hold
in the highest regard and I will cherish it forever. I thank him for being so patient with me

all through these years and taking care of me both at a professional and a personal level.

iv

To my Committee Members, Prof Timothy Mrozowski and Dr. Mark Meerschaert,
thanks to them for serving as co-chairs and playing the devil’s advocate when it was
required and giving me valuable inputs to help refine my thesis. Also, I thank them for

making classroom-learning a highly enjoyable experience in their respective expertise.

A special note of thanks to Mr. Steven J. Pierce and Dr. T enko Raykov, for their expert

guidance towards the end of my research, they made climbing the last mountain easier.

I would like extend my special thanks to Dr. Matt Syal, for his advices on matters of
master’s program and career search. His energy levels are unmatched for a person who
has so much on his plate, a source of inspiration for how to tackle the busy life that is

dawning upon me.

I would like to extend my cordial thanks to Dr. Mohamed El-Gajy, for making classroom
learning exciting, connecting with students at a personal level and he always inspires

students to push the envelope.

I extend my gratitude to all the administrative staff and faculty members, with special
mention to Ms. Cathy Morrison, who by their sheer presence, made college matters

between any two classrooms stress-free.

I would like to express my appreciation and warmest regard for my colleague and friend
Don Schafer, a man of few words but lots of wisdom and ideas. It was a pleasure to share
the research laboratory with him. It was a discussion with him that helped me come up

with this thesis idea, and his logistical support at crucial times helped my research.

To my parents, who are my pillars of strength and motivation, I offer my deepest respect
and gratitude for encouraging and supporting me to pursue my dreams and showering all

their blessings on me.

To my brother, Saurabh Jain, I offer my deepest love and regard for letting me selfishly
pursue my dreams and taking on my responsibilities within the family. His sacrifice is the

hardest and I will be forever indebted. You are the best.

A special note of thanks and regards to my uncle and aunt, Mr. Anil Agarwal and Mrs.

Rachna A garwal, for encouraging me to pursue my dreams.

Last but definitely not the least, I express my warmest hugs and give high-fives to all my
fi’iends, who I had the pleasure to know from my undergrad days, Amanjeet Singh, Yash
Pratap Singh, and Shilpi Mago, they made sure that I stay focused in my endeavor, and
friends who I met here for the first time on Spartan soil with special mention to Lipika
Swamp, Sagata Bhawani, Nikhil Mehrotra, Radhika Arora, Pranav Sané, Rajat Basu, Ali
Lahouti and Alessandro Orsi for giving me a family away fi'om home. It is a testament to
your spirit and cheerful company that tough times were so easily tackled, and fun times

never lefi my side. Cheers to all of you.

Finally, thanks to all those friends and colleagues who I have not mentioned but they
have a special place in my heart and memory, my interactions with you have left positive
imprints on me that helped me shape my opinion about this multicultural world and help
me grow as a responsible person. I hope to stay in touch with all of you and continually

affect each other’s lives in a positive manner.

...... I did it!

vi

LIST OF TABLES .................................................................................................... x
LIST OF FIGURES .................................................................................................. xii
CHAPTER 1
INTRODUCTION
1 . 1 Introduction ............................................................................ 2
1.2 Need Statement ....................................................................... 5
1.3 Research Questions ................................................................... 7
1.4 Research Goals, Objectives and Methods ........................................ 8
1.5 Research Benefits and Contribution ............................................... 9
1.6 Chapter Summary .................................................................... 10
CHAPTER 2
LITERATURE REVIEW
2. 1 Introduction .......................................................................... 12
2.2 Production Management and Control ............................................. 12
2.3 Current Production Management and Production Control in Construction.. 15
2.4 Lean Production Management ...................................................... 24
2.5 Production factors of Construction Projects ...................................... 39
2.5.1 Classification of Production Factors ..................................................... 40
2.6 Structural Equation Modeling: An Introduction ................................. 43
2.6.1 Basics of SEM Estimation and Assessment ................................ 48
2.6.2 Components of Structural Equation Modeling ............................. 50
2.6.2.1 Development of Measurement Model .................................... 59
2.6.2.2 Sample Size in S.E.M ............................................................. 61
2.6.2.3 Development of Structural Model .......................................... 65
2.7 Need and Application of SEM in Construction..................................... 68
2.7.1 Current Trends in Application of SEM Technique ............................... 72
2.8 Chapter Summary ..................................................................... 75
CHAPTER 3
RESEARCH METHODOLOGY
3. 1 Introduction ........................................................................... 7 7
3.2 Existing Literature, Scope of Development and Research Contribution. 82
3.2.1 Phase-1 Literature Analysis .................................................................. 82
3.2.1.1 Conventional Production Management Techniques .............. 83
3.2.1.2 Lean Construction Management Techniques ......................... 84
3.2.1.3 Structural Equation Modeling ................................................ 85

TABLE OF CONTENTS

vii

3.2.2 Phase-2 Establish Scope of Development ............................................. 87
3.2.3 Phase—3 Develop Framework for designing Causality Equation ........... 87
3.2.3.1 Framework Demonstration ..................................................... 88
3.4 Chapter Summary ........................................................................................... 89
CHAPTER 4
FRAMEWORK DEVELOPMENT AND DEMONSTRATION
4. 1 Introduction ................................................................................................... 9 l
4.2 Scope of Development of Framework ........................................................... 91
4.3 Framework Development .............................................................................. 92
4.3.1 Phase-1 State Hypothesis ...................................................................... 92
4.3.2 Phase-2 Identify the Constructs ............................................................ 93
4.3.3 Phase-3 Define and Operationalize the Individual Constructs ............. 96
4.3.4 Phase-4 Construct Validation ............................................................... 105
4.3.4.1 Develop the Overall Measurement Model ............................ 107
4.3.5 Phase-5 Develop and Specify the Structural Model ............................. 111
4.3.5.1 Model Identification ............................................................. 114
4.3.6 Phase—6 Design Study to Produce Empirical Results ........................... 117
4.3.6.1 Determining Sample Size ...................................................... 117
4.3.6.2 Handling Missing Data ......................................................... 119
4.3.7 Phase-7 Data Collection ........................................................................ 120
4.3.7.1 Multi-Collinearity .................................................................. 121
4.3.8 Phase-8 Assess Structural Model Validity ............................................ 122
4.3.9 Phase-9 Interpreting the Results ........................................................... 126
4.4 Discussion of Strategies to Assist in Model Development ............................ 135
4.5 Developed Framework ................................................................................... 138
4.6 Chapter Summary .......................................................................................... 142
CHAPTER 5
SUMMARY AND CONCLUSIONS
5. 1 Introduction ................................................................................................... 144
5.2 Research Overview ........................................................................................ 144
5.3 Research Goals and Objectives ...................................................................... 146
5.4 Conclusions and Inferences .......................................................................... 147
5.5 Limitations of the Research ........................................................................... 150
5.6 Research Benefits and Contribution .............................................................. 151
5.7 Future Areas of Research ............................................................................... 155
5.8 Chapter Summary .......................................................................................... 157
APPENDICES

viii

APPENDIX A Survey Instrument .................................................................. 159
APPENDIX B Output File from EQS for Framework Demonstration

Example ................................................................................. 172
APPENDIX C Source of each question in the checklist ................................ 254
REFERENCES ........................................................................................................ 259

ix

Table 2.1

Table 2.2

Table 2.3

Table 2.4

Table 2. 5

Table 2.6

Table 2.7

Table 4.1

Table 4.2

Table 4.3

Table 4.4

Table 4.5

Table 4.6

Table 4.7

Table 4.8

Table 4.9

Table 4.10

Table 4.11

Table 4.12

LIST OF TABLES

TFV Theory of Production ........................................................................... 14
Interpretation of Variance ............................................................................ 17
Proposed PWR Metrics ............................................................................... 38
Classification of production factors suited to Last Planner System ............ 43
Observed Covariance Matrix for model in Figure 2.13 .............................. 49
Conceptual differences between Formative and Reflective Indicators ....... 56
Statistical differences between Formative and Reflective Indicators .......... 57

Categorizing the constructs into endogenous and exogenous constructs
(with abbreviation) ...................................................................................... 95

Measured Variables for Work flow Reliability Construct .......................... 98

Measured Variables for Availability of Pre Requisite Work Construct ...... 99

Measured Variables for Availability of Directives Construct ................... 100
Measured Variables for Availability of Resources Construct ................... 101
Measured Variables for Waste Reduction Construct ................................ 102
Measured Variables for Variation Reduction Construct ........................... 103
Measured Variables for Burden Reduction Construct ............................... 105
Multi-collinearity test for Prerequisite Construct using MINITAB .......... 122
Command file generated after selecting estimation method .................... 124
Goodness-of-Fit Statistics for the demonstration model .......................... 126

Characteristics of different Fit Indices demonstrating
Goodness-of—F it across different model situations .................................. 127

X

Table 4.13 Covariance Matrix (trimmed) and model specification
& identification results ............................................................................. 129

Table 4.14 Standardized Residual matrix (Trimmed) and largest
standardized residuals ............................................................................... 131

Table 4.15 Standard Errors, Variances of Independent Variables,
and Standardized Solution (Edited) ......................................................... 133

Table 5.1 Goodness-of—fit Results for waste-variation-burden model ..................... 153

Table 5.2 Comparison of parameter estimates (Standardized solution)
of same variables in two models .............................................................. 154

xi

Figure 2.1
Figure 2.2
Figure 2.3

Figure 2.4

Figure 2.5

LIST OF FIGURES

Variance Analysis ....................................................................................... 17
Example of Line of Balance Scheduling Method ....................................... 21
‘Last Planner System’ (with additions in red) .............................................. 26

Improving the trade-off between time and cost by reducing workflow
variation ...................................................................................................... 29

Constraints Analysis sheet .......................................................................... 30

Figure 2.6 Pie Chart showing frequencies of various reasons for incomplete work

Figure 2.7
Figure 2.8
Figure 2.9
Figure 2.10
Figure 2.11
Figure 2.12
Figure 2.13
Figure 2.14
Figure 2.15
Figure 2.16
Figure 3.1
Figure 3.2

Figure 3.3

(<100% PPC) .............................................................................................. 31
Elements and Metrics of Make Ready Process ........................................... 33
Planned Work Ready for lookahead period ................................................ 34
Eamable Value ............................................................................................ 35
Time-Time chart showing work forecast and work performed .................. 36
Sketch showing different components of a structural equation model ....... 46
Sketch showing exogenous and endogenous constructs ............................. 47
Sketch showing Covariances that are calculated in a structural model ...... 48
Sketch showing reflective construct and indicators .................................... 53
Sketch showing formative construct and indicators ................................... 55
Sketch showing a MIMIC model ................................................................ 58
Research Methodology in 4 steps ................................................................ 78
Detailed research methodology .................................................................. 80
Five basic steps of designing a Structural Equation Model ....................... 86

xii

Figure 3.4

Figure 4.1
Figure 4.2
Figure 4.3
Figure 4.4
Figure 4.5

Figure 4.6

Figure 4.7

Sample representation of final framework ................................................. 88
Path Diagram of the proposed structural model (lSt Stage) ........................ 96
Diagram showing the constructs in a measurement model relationship 108

Diagram showing the Structural Model of the proposed Hypotheses ....... 113

Structural Model showing fixed and free parameters. .............................. 116
EQS window showing selection of Estimation method ............................ 123
Window showing Output options for Model Estimation Results .............. 125

Developed Framework for Assessment of Cause-and-Effect
Relationship using SEM ............................................................................ 138

xiii

CHAPTER-1

INTRODUCTION

Small changes in the starting conditions in a complex dynamic system produce outcomes
totally out of proportion to their magnitude, making the phenomenon or system inherently

unpredictable in the long term. — Chaos T heory1

1.1 Introduction

Many factors contribute to a project’s success, but recent research points to honoring
contractual commitments, open and effective communication of team members, and clear
project goals and objectives as very critical factors for success. At the production level,
improving work flow reliability between production units is of paramount importance,
perhaps even exceeding the importance of increasing the productivity of any give unit. A
major contributor to improving work flow reliability has been the explicit application of
production management techniques as inspired by Lean Construction. A key component
in addressing work flow reliability is comparison between work planned and work
performed on a weekly and even on a daily level.

Since the 19605, construction management practices were mainly directed at the
project planning and control levels, which have improved the ability to oversee and
manage construction projects at a macro level but not at the site production level. Tools
such as Critical Path Method (CPM) and Gantt Bar Chart are used for scheduling of
construction activities as well as monitoring progress of the project in terms of duration
by comparing the original schedule to the working schedule. The Earned Value Method
(EVM) is another project control tool which provides a comparison between scheduled

costs and actual costs expended over the duration of a project and assist in estimating the

 

l . . . .. . .
hupzllwmvbusmessdrctrogarycom/detinitron/chaos-theoryhtml, adopted from several other sources as
well.

1.: gens-'—

financial health of the project. Other method such as Line-of—Balance Scheduling (LOB)
method, which were developed in the ‘503 and faded in the ‘705, has been made popular
again in the 90’s. LOB techniques are primarily implemented for projects with repetitive
units.

Although, the project management methods and tools mentioned above have
proven to be excellent tools in reviewing the status of projects at any given point of time,
they haven’t been able to provide a total control to the management team over the events
that take place in the field. The CPM and EVM are used together and implement the
‘percent work complete’ to assess the situation of the project. The former is used to
compare the stipulated duration vs. the actual duration of the project and the latter is used
to compare the actual expenditure vs. the budgeted expenditure of the project. In both
methods, the information available to the management team pertains to the current status
of the project and provides no indication about the status of the upcoming work.
Therefore, the management team can only take corrective actions for the problems that
have occurred whereas no action can be taken in time to prevent any firture problems
from occurrence. This lack of control has handicapped the construction management
teams in exercising a command over projects and has been a contributing factor to low
workflow reliability; where workflow is the progression of work within an activity and
between activities.

Improving the workflow reliability has been the primary goal of Lean
Construction theory and practice. In order to achieve this goal, researchers in this field
developed advanced techniques like lookahead planning to attain control over the tasks

and activities performed at the jobsite. The purpose of the lookahead planning is to plan

out the work to be performed in advance for the upcoming weeks (up to 6), and compare
the performed work with the planned once the work week is over. This measurement is
termed as “percent plan complete (PPC)” and indicates the shortcoming of the production
planning on site with the help of reasons analysis. PPC is a better tool in comparison to
“percent work complete” of EVM and CPM because the former compares the WILL vs.
DID work whereas the latter compares the SHOULD vs. DID work (Mitropoulos 2005).
This makes the PPC a more accurate source of measurement since it compares only the
work (that WILL be done) reflecting the true production capability of the construction
team to the actual performed work instead of the work (that SHOULD be done) that has
piled up due to previous interferences. However, similar to “percent work complete”,
PPC is also a lagging indicator, and helps in surfacing the problems only after they have
occurred. A reasons analysis conducted after measuring PPC identifies the reasons only
after work stopped. This makes the nature of the solutions as remedial in nature rather
than preventive. PPC in its current state is not a predictor of whether the work will be
executed as planned.

Although research contributions have added ‘planned work ready’ (PWR) metric
to gauge whether the work planned stays the same through the preceding six weeks and
enters as such into the actual work week, it helps the planner to remove any constraints
identified during the six week period but does not warn them about any hidden
constraints or variation of a removed constraint which may lead it to resurface during the
actual work week. It may be appropriate to mention that research on PPC at this stage is

focused on ways to improve it by means of other support metrics; however, there is no

I~J

research that attempts to define PPC in terms of its relationship with the constraints that

either cause the work to be executed or not.

Control means causing a desired future rather than identifying variances between plan
and actual (Ballard 2000). Clearly, none of the project and production metrics provide the
construction manager a control over the project in the context stated. CPM and EVM are
only performance indicators of a project and do not diagnose the problem of the project,
whereas PPC is more advanced and does help in correcting the problems through reasons
analysis. The use of PPC has improved project controls (PPC has reached 90% in some
cases (Ballard 2000)) but it is still short of providing 100% reliability over the week to
week progress of work. There is a research need to understand, in a causal fashion, what
factors contribute to reliable workflow so that these factors are addressed before the plan
is executed. The aim of this research is to provide project participants with the ability to
improve understanding of factors affecting workflow so that their objectives are met in

time, cost and quality as agreed.

1.2 Need Statement

The Last Planner ® System uses PPC as a measure to track the progress of a project. PPC
essentially helps measure the extent to which production planning results in reliable work
flow, quantitatively. The purpose of PPC is to help compare the actual vs. planned work
during a week and identify the causes of incomplete work, but this planned work is result
of commitments made by the contractor, therefore, the PPC checks the true reliability of
the production planning. Using PPC has an advantage over traditional project control

tools like Earned Value Method and Critical Path Method because, PPC along with Last

planner, in metaphorical terms, prescribes healthy behavior so that the body is ill-free,
whereas the EVM and the CPM method only help in taking the temperature of the
project. Another analogy from the financial sector, EVM and CPM are like Dow Jones
Industrial average in that they indicate the general health of the market but the status of
an individual company cannot be assessed just by looking at the Dow, also it cannot be
inferred whether the company’s stock value will go up or down the following day just by
looking at the Dow value.

From a lean production perspective, schedule controls have been criticized for
their inability to control and stabilize the workflow (Mitropoulos 2005). Kim and Ballard
(2000) mentioned that EVM does not consider the work sequence, and as a result,
managerial actions driven by EV typically increase the variation of workflow. In
addition, the traditional control systems do not provide any indication about the status of
upcoming work, i.e., how much of the upcoming planned work can be performed as
planned, hence it limits their ability to control the project progress, and their ability to
identify appropriate corrective action before problems are encountered (Mitropoulos
2005)

Jobsite management has different levels of control on the factors affecting work
flow reliability. Some factors, such as material and tool availability, can be managed and
controlled. Other factors, such as the availability of skilled craft workers and extreme
weather, are difficult to eliminate even though various methods are available to minimize
their impact (Dai et al. 2009b). The development of look ahead process and tracking of
PPC in this light, have made great strides in production planning by improving the ability

to increase the control over project progress; Ballard (2000) reported that with improved

implementation of lookahead process the PPC achieved rose from 70% to 80-85% in
some cases and above 90% in one case. Although PPC helps in identifying the problems
and suggest corrective actions, it comes into action only after occurrence of the
problem(s). To ensure 100% reliability in work planning, the planner must understand the
relationship between PPC and the production factors, which is currently lacking in
practice and research. It is widely believed that the majority of the factors affecting
construction productivity can be improved through the efforts of jobsite management
(Dai et al. 2009b). An understanding of the relationship between production factors and
their effect on project work flow reliability will help identify in advance the potential
problems that could affect the work performance during the work week and will better
equip the planners to strive for a 100% PPC on a consistent basis.

Mitropoulos (2005) developed make ready metrics to improve upon the lookahead
process and add the capacity to measure the accuracy of the forecast of PWR metric; it
proposes to check the accuracy of forecast of work planned six weeks earlier vs. every
progressive week (every subsequent lookahead horizon). Integrated with the lookahead
planning and post-work analysis process, PWR provides the best opportunity to establish
a causality model with the production factors which would help attain the ideal of 100%
work flow reliability in a project. The identification of the need to develop methods to
establish relationship between production factors and the PPC in order to concentrate on
the planning efforts before work week defines the goal and objectives of this research, as
decisions taken during the planning process have been found to have a significant

influence on the probable outcome of the project (Arditi 1985).

1.3 Research Questions
The lack of an explanatory understanding of factors giving rise to low/high PPC
highlights the importance of investigating the current production metrics implemented in
Last Planner System and how these metrics affect workflow reliability in the production
stage. As production is cumulative in nature, where underperformances and deficiencies
multiply as we move downstream it is important to investigate the combined effect of
constraints and underperformances on the next production performance output, and
develop an effect-cause-effect relationship between production constraints and percent
plan complete. This research poses the following questions:
1. What current ability exists to forecast production performance using the Last
Planner system?
2. What factors give rise to reliable workflow? How can we use those factors to
predict work flow in measure of PPC?

3. What method can be implemented to study these factors?

1.4 Research Goals, Objectives, and Methods
The overall goal of this research is to understand cause and effect relations related to
workflow reliability at the production level in a construction project. This goal can be
approached in many ways, albeit not fully reachable.

To approach this goal, the research will focus on developing a method to
investigate the causal relationship between production constraints and their impact on
workflow reliability as measured by the PPC metric. To accomplish this, the following

objectives are proposed:

1. Study production management tools implemented at the site level and document the
production delay factors encountered on construction site.
Objective 1 is fulfilled by performing following steps:

a. Conduct literature review on production planning and control tools,
productivity measurement studies and identify factors causing work flow
variability.

2. Develop a method to study the relationship between production delay factors and
work flow reliability.
Objective 2 is fulfilled by performing following steps:

3. Develop a Survey Instrument to measure the existence of identified
production factors.

b. Based on the literature review, develop a framework to design a causality

model between various production factors and work flow reliability.

1.5 Research Benefits and Contribution
This research will develop a method to assess the impacts of production factors on the
workflow on a construction site. This method would be implemented by the production
management team to help identify and remove factors that could result in unreliable
workflow.

The research will contribute to the project and production management practices
for construction projects by developing a method to better understand the cause and
effect of factors influencing reliable workflow. The framework could be adopted to

assess impacts of project planning factors at various stages of a construction project;

which will aid different participants in identifying factors and analyzing the collective
impact of those factors on the work to be executed.

The research will mainly benefit contractors and subcontractors who implement
the Last Planner System in their projects by aiding them in improving their understanding
of factors impacting reliable workflow as well as the ability to conduct long term studies

by using the survey instrument.

1.6 Chapter Summary

Section one discussed the current level of implementation and understanding of PPC for
project control purposes and highlights the need to improve it to achieve 100% work flow
reliability. The research goals and objectives were identified that would help in achieving
the desired outputs from the research. The research aims to aid the construction
production planners by developing a method that will enable them to understand the

impact of various factors on the effectiveness of their weekly work planning.

CHAPTER-2

LITERATURE REVIEW

2.1 Introduction

This chapter discusses the existing literature of production management practices and
tools used on construction projects. The purpose of reviewing production management
practices is to understand the extent of control sought at the management level, what is
the level of predictability possible in assessing the progress of the project, how accurate
control tools act as indicators of project performance.

The terms ‘production management’ and ‘production control’ are used
interchangeably in the context of construction industry; this is mainly due to the fact that
the production management practices are often identified by the production control tools.
Therefore, the literature review of both has been combined to facilitate explanation and
discussion of the literature background.

The study also aims to review the production factors/constraints that make a
project plan fail to achieve the desired performance and whether it is possible to control
those factors if identified early. The aforementioned literature review helped in
addressing the research questions and enriching the focus of achieving the research

objectives.

2.2 Production Management and Control

“Production has three kinds of goal. First, there is the goal of getting intended products
produced in general. Second, there are goals related to the characteristics of the
production itself, such as cost minimization and level of utilization (internal goals).
Third, there are goals related to the needs of the customer, such as quality, dependability

and flexibility (external goals)” (Koskela 1999). These were the basis to form three

 

different views of production; transformation view, flow view, and value-generation
view. The transformation view is instrumental in discovering which tasks are needed in a
production undertaking and in getting it realized. The flow view focuses on eliminating
wasteful processes that delay and interrupt work flow. In value generation view, the basic
goal is to reach the best possible value from the point of the customer.

Production has been an explicit topic of study mainly in industrial engineering,
which has dealt mostly with one type of production; namely, manufacturing with only
occasional forays into construction industry (Ballard 2000). The term “manufacturing” is
most commonly used to describe the making of many copies from a single design, and is
primarily focused on making products for the mass market, which in most cases the
products are moveable within the assembly line or in certain cases not moveable like
ships or airplanes (Ballard 2000). With this understanding of manufacturing,
“manufactured housing” is the only point of coincidence between construction industry
and the term production (Ballard 2000), without which construction industry is largely
seen as a service based industry with designing and engineering viewed as services and
not products.

Defining production as the designing, engineering and making of artifacts allow
us to understand how construction is a type of production and how design is an essential
component in construction (Ballard 2000).

The three views of production management had their inefficiencies when
implemented in isolation; therefore, to overcome these problems, a theory of production
that incorporates transformation, flow, and value view of production was introduced.

Koskela (1992) formulated a TFV theory of production which integrated the

-13-

 

transformation, flow, and value views. The new theory was advanced on the basis that the

three views of production do not present alternative, competing theories of production,

but rather theories that are partial and complementary. Table 2.1 below indicates the

components of the TFV theory.

 

 

 

 

 

 

 

Transformation Flow View Value Generation
View View
Conceptualization As a transformation As a flow of As a process where
of production of inputs into material, composed value for the
outputs of transformation, customer is created
inspection, moving through fillfillment
and waiting of his requirements
Main principle Getting production Elimination of Elimination of value
realized efficiently waste (non-value loss (achieved value
adding activities) in relation to best
possible value)
Methods and Work breakdown Continuous flow, Methods for
practices structure, MRP, pull production requirement capture,
Organizational control, continuous Quality Function
ResponsibilityChart improvement Deployment
Practical Taking care of what Taking care of that Taking care of that
contribution has to be done what is unnecessary customer
is done as little as requirements are
possible met in the best
possible manner
Suggested name for Task Management Flow Management Value Management
practical
application of the
view

 

 

 

 

Table 2.1 TFV Theory of Production (Source: Koskela 1992)

The term ‘control’ as found in Concise Oxford Dictionary, means to dominate,

command; to check, verify; to regulate. In reference to the project control theory, control

essentially means to keep an account of things, the main purpose being to monitor actual

costs and schedule performance against target in order to identify negative variances

(Ballard 2000). This is considered reactive, and a view of control as “making things

happen” has been advocated.

 

Production control theorists working in manufacturing distinguish two primary
ways of regulating work flow in manufacturing systems: push and pull. Push systems
release materials or information into a system based on pre-assigned due dates (from a
master production schedule, for example) for the products of which they are parts
(Ballard 2000). Pull systems release materials or information into a system based on the
state of the system (the amount of work in process, the quality of available assignments,
etc) in addition to due dates (Hopp and Spearman 1996). In factory systems, pull should
be driven ultimately from customer orders. In construction, pull is ultimately driven by
target completion dates, but specifically applies to the internal customer of each process
(Ballard 2000).

The traditional project control systems implement the push system, which does
not always produce the desired results from the project. The limitations of different

current production management practices will be discussed in the following section.

2.3 Current Production Management and Production Control in Construction

Project control tools are commonly used in the construction industry; unfortunately,
many projects run over budget and behind schedule, which suggests that there is
something wrong in our project control system (Kim and Ballard 2000). Construction
projects are managed today by breaking them into pieces or activities, estimating the time
and money to complete each, applying the critical-path method (CPM) (there are many
other methods of scheduling that are deployed in construction but CPM is by far the most
popular method) to identify a logical order, and then either contracting externally or

assigning internally to establish responsibility. Project managers use the schedule to

determine when each activity should start and push for work to being on the earliest start
date (Ballard et a1 2002). Time and cost targets are monitored on a pre-established cycle
and actions taken when off-target measures result. This traditional system of project
control is also known as the thermostat model of control (Moder, Phillips and Davis
1983). The thermostat model triggers action when a variance is detected and it assumes
that there are direct links to the cause of the variance (Ballard et al 2002). Conventional
project management in construction is inadequate because it does not rest on a TF V
theoretical framework (Howell and Koskela 2000).

Projects today are complex, uncertain and quick (CUQ) (Shenhar and Laufer,
1995). Co-ordination of work on CUQ projects cannot be assured even with highly
detailed CPM schedules because these schedules portray the project as a series of
activities and ignore the flow of material and information within and between them. The
reliable release of work from one crew to the next is assumed or ignored (Ballard et a1
2002)

Controlling of projects using CPM scheduling is achieved using the Earned Value
Method (EVM), both these techniques are implemented in tandem. The CPM compares
the actual progress with the baseline schedule, and monitors the time floats on the critical
and near critical activities. The EV method monitors the progress of activities using
dollar value as the metric, that is, by comparing the Budgeted Cost of Work Scheduled
(BCWS) with the Budgeted Cost of Work Performed (BCWP). For each activity, the
schedule variance (SV) is calculated as the difference BCWP — BCWS. The cost

variance, for each activity, is calculated as the difference Budgeted Cost of Work

Performed (BCWP) — Actual Cost of Work Performed (ACWP). The graphical
illustration of the SV and CV calculation is shown in the Figure 2.1.

The project progress is indicated by aggregating the SV and CV values of the
individual activities. If the $ value of the work performed (upto the reporting date) is
more than the work scheduled and cost less than budgeted, the project is considered

‘ahead of schedule’ and ‘cost underrun’ (refer Table 2.2) (Mitropoulos 2005).

 

 

 

 

 

 

 

 

 

 

Report date
Budget at completion
0
_ BCWS :3
O.
__ E
e 23
a — ac)
_ I AV 0.
CV
‘ ACWP j sv
BCWP
l I l l l l l l I 1
Reporting Period
Figure 2.1 Variance Analysis (Source: Kim and Ballard 2000)
Variance - 0 +
Cost Variance (CV) Cost Overrun On Budget Cost Underrun
Schedule Variance (SV) Behind Schedule On Schedule Ahead of Schedule

 

 

 

 

 

 

Table 2.2 Interpretation of Variance (Source: Kim and Ballard 2000)

Though, the EVM is considered the most advanced technique for integration of
schedule and cost (Kim and Ballard 2000), researchers have criticized the EVM approach

for a variety of reasons. Mitropoulos (2005) mentioned that CPM and EVM have a

limited ability to control the project progress because they do not provide any indication
about the status of the upcoming work, except through predictions, this further limit their
ability to identify appropriate corrective action before problems are encountered. This
may lead to a situation where a project that is ahead of schedule but has only a small
portion of the upcoming planned work that can be performed may not be considered
problematic. Conversely, a project that is behind schedule but has a lot of upcoming work
that can be performed may be considered a lot more problematic. In addition, corrective
action taken based on progress to date may compound the project problems if they do not
consider the status of the upcoming work (Mitropoulos 2005).

Walt (2006) criticized Earned Value Method for being only a cost management
tool stating that information relating to schedule performance is inadequate. Walt (2006)
outlined three major deficiencies in the EVM method:

1. The quality performance indicators are not directly connected to the project
output. For example, milestone completion or delivery of products may not meet
the customer’s expectation, yet EVM indicate acceptable values.

2. The schedule indicators are flawed. For projects completing late, the indicators
always show perfect schedule performance (which is due to the fact that EVM is
not designed to indicate performance related to time at all).

3. The performance indicators are not explicitly connected to appropriate
management action. Even with EVM data, the project manager remains reliant on
intuition as to any action needed.

Kim and Ballard (2000) described the EV method as a project control technique

which provides a quantitative measure of work performance which involves a crediting of

budget dollars or labor hours as scheduled work is performed and credited it as a superior
control technique since it integrates schedule and cost performance of the project. At the
same time, they addressed vulnerability in its conceptual framework, which, stands on the
assumption that one earned hour is as good as another, and the correlative assumption
that the productivity of each type of work activity is independent of the performance of
other work activities, even when they are in predecessors-successor network.

The EVM technique uses cost accounts as management control points because
they are the lowest level at which individual variance analysis can be made. The US
Department of Defense (US DOD) describes a cost account as a natural control point for
cost/schedule planning and control since it represents the work assigned to one
responsible organizational element on one contract work breakdown structure element.
Since integrated control of cost and schedule is core in EVM, it is desirable that cost
accounts be identical to elements activities in the network schedule. But since cost
accounts are too coarse to be assigned in schedule, each account has work packages with
their own schedule durations and assigned budget (Kim and Ballard 2000).

Notably, if schedule variance is shown as a negative value on a specific cost
account as of a reporting date, the manager of the ‘red flagged’ cost account gets in
trouble. The manager attempts to remove this problem by increasing the earned value
(BCWP) of his cost account as much as possible. Since detailed work procedure and/or
sequence is usually at the manager’s discretion, managers manipulate work sequences or
release work assignments in order to make their performance better, without regard to

work flow uncertainty and its negative impact downstream (Kim and Ballard 2000).

Kim and Ballard (2000) also stated that EVM does not differentiate between
value-generating operations and non-value generation because calculation of BCWP
disregards downstream demand. They attributed this to the lack of ability of EVM
method to recognize the couplings that exist between accounts or activities: intermediate
product and shared resources. Based on their research, Kim and Ballard (2000) concluded
that making decisions for releasing work/taking corrective action on basis of cost account
progress result in longer durations and higher costs than necessary. They proposed five
criteria for generating quality assignments/releasing work: Definition; Soundness;
Sequence; Size; and Learning. These five criteria would form an essential part of the
lookahead planning process and for releasing quality work for a scheduled week (will be
discussed later in this chapter).

Another, less widely used production management method in construction is the
Line of Balance (LOB). Originally, developed by Goodyear Company as a linear
technique in early 1940’s, and was later developed by National Building Agency (UK)
for repetitive housing projects, where a resource-oriented scheduling tool — that
considered resources as a starting point —- was considered more appropriate than the ones
which are activity—dominated, as in CPM. The Line of Balance works on the premise that
activities should be planned according to their production rhythms, in other words, the
number of units that a crew can produce in a determined time unit. These rhythms are
represented in a graphical format which shows clearly the production rates of the various

activities against time (Henrich and Koskela 2006) (refer Figure 2.2).

-20-

         
    
   

IUII'L'YI" I r ' IDL’EI‘

N wptoiocl;
Line-of Balance n' m,

[I “my profit '

. Ordero
completion
(*ka J ‘
(mal’ .

' 77/

:qu Moo)

 

 

.1" /1

Work flow 1

 

 

 

 

 

       

 

 

 

 

 

 

....... f Space buffer S:
l
’ a
l
1 1
Plan- Irwin—Lina ----.-- --- 91.3mm ._ _. _ i
lesion Inoce
uruarropct v1.11: 7
—‘J I f

 

 

 

 

7 Time

Figure 2.2 Example of Line of Balance Scheduling Method (Source: Kankainen and Seppanen 2003)
The LOB method provides a simple visual tool that helps the manager of a

process to observe the progress of each activity. This assists the manager in making
decisions such as: level of detail in activities planning, crew size, production expected
and achieved, production rhythm and learning. This in turn helps the manager determine
the number of crew simultaneously on the site, their position and location, the direction
of production, and equipment available or able to be used (Henrich and Koskela 2006).
Some advantages of using LOB scheduling in the context of schedule planning as
documented by Kankainen and Seppanen (2003) are; 1) it gives a better control of work
groups, since they form the basis of planning activities, 2) allows for an integrated
procurement schedule very early with master schedule so constraints on material
availability, labor, contracts, and engineering are taken into account, 3) buffers can be
planned to minimize the effect of work flow variability allowing production to be

implemented as planned provided an adequate space buffer has been planned between

-21-

space-critical tasks, 4) increasing productivity, less waiting hours and less hurried work,
5) use of resources can be planned to be continuous and level which results in lower costs
and less deviations in production.

A certain advantage of LOB scheduling over CPM is that the production rates and
duration information is represented in a graphic format in LOB while CPM does not give
any indications. The LOB plot can show at a glance what is wrong with the progress of
an activity, and can detect potential future bottlenecks. Obviously, LOB allows a better
grasp of a project composed of repetitive activities than any other scheduling technique,
because it allows the possibility to adjust activities’ rates of production (Arditi et a1
2002). LOB is oriented toward the required delivery of completed units and is based on
knowledge of how many units must be completed on any day so that the programmed
delivery of units can be achieved. Once a target rate of delivery has been established for
the project, the rate of production of each and every activity is expected not to be less
than this target rate of delivery (Lumsden 1968). The optimum rate of output that a crew
of optimum size will be able to produce is called the “natural rhythm” of the activity
(Arditi et a1 2002). This natural rhythm is difficult to achieve in a non-repetitive project,
therefore, the LOB method is less successful in such category of projects, or even phases
of the same project such as foundation work.

While, it was noted as an advantage that buffers can be planned to minimize the
effect of work flow variability (Kankainen and Seppanen 2003), the space buffers
become a function of the productivity rate available to the company in their records based
on their previous projects and there is no conscious consideration of constraints while

deciding the duration of a particular project. The repetitive activities are treated with

-22-

respect but the non-repetitive and discrete activities are usually clubbed after the first
calculation of project duration, according to the precedence relationship with repetitive
activities. The discrete activities are handled as workable backlog in case of any forecast
of a possible bottleneck.

Again, the adjustment of productivity rates of adjacent activities or space-critical
activities on the basis of time required in implementing a control-action to overcome a
bottleneck or any other problem represents a gambling situation instead of an informed
approach. The LOB method is trusted on the assumption that the natural rhythm of the
activity will hold and space buffers and time buffers are added on the basis of previous
experience and number of discrete activities. It reflects the lack of ‘how to achieve the
work’ approach of the weekly work planner in the Last Planner System. Needlessto say,
the bottlenecks surface as a result of something not gone exactly as planned and
identification of constraints alter the execution stage. The graphical format of LOB may
help in identifying the bottlenecks in fiiture work but only once a current activity is left
incomplete due to unidentifiable constraints, a sign of inefficient project control plan.

Clearly, like CPM, the LOB method also presents itself as a production
management tool that is dependent heavily on past experiences and does not have a crash
proof solutions kit that can confidently keep the progress under control and prevent
spiraling of unforeseen problems. It separates itself from CPM on scheduling principles,
i.e., CPM is an Activity-based scheduling based method whereas LOB is a resource based
scheduling method but principally both methods are handicapped when it comes to
prepare a production plan according to the true capacity of the construction team

(currently, it is prepared on the assumed capacity of the construction team).

-23-

The need for an effective production control without the previously discussed
shortcomings was identified in the Lean Construction philosophy during the 1990’s.
Research in this field has led to developments such as LPDS (Lean Project Delivery
System) and LPS (Last Planner® System) which are positive developments in the desired
direction. The use of lean tools in production control will be discussed in the following

section.

2.4 Lean Production Management

Project and Production management is at the heart of Lean Construction and runs from
the very beginning of a project to handover of a facility to the client. Lean project and
production management is accomplished using Work Structuring and Production Control.
Lean work structuring is process design integrated with product design and extends in
scope from an entire production system down to the operations performed on materials
and information within that system. It produces a range of outputs such as project
execution strategies, project organizational structures, operations designs, master
schedules, and phase schedules (Ballard et al 2002). As envisioned, LWS begins during
the schematic design phase and continues to construction pre-planning.

Production control governs execution of plans and extends throughout a project.
Production control consists of work flow control and production unit control. Work flow
control is accomplished primarily through the lookahead process. Production unit control
is accomplished primarily through weekly work planning (Ballard 2000). The Lean
Construction Institute (LCI) developed Last Planner® System (LPS) (Figure 2.3) as a

production control tool. The purpose of LPS is to tie up the work structuring at the front

-24-

end planning stage with the actual production at site with a systematic and uninterrupted
work flow while generating maximum value to the customer. The LPS is based on the
Lean principles of:
o reaching consensus in all decision making
0 using set-based design approach (both of these are realized in work structuring
stage)
0 doing a genchi genbutsu" on project site and develop lookahead plans and weekly
work plans and assign tasks to workable backlog after doing a constraint analysis
0 And emphasize on continuous improvement by the measure of charting percent
plan complete (PPC) of the work and analysis of reasons for unaccomplished

tasks to prevent a repetitive error.

 

2 Defined as “Go and see for yourself to thoroughly understand the situation”. Liker J. (2004)

-25-

 

 

LAST PLANNER SYSTEM

 

 

n

( esign
Criteria

 

SHOULD

   
     
   

 

tr

 

   
 
  

 

 

 

 

 

Master
& Phase 0
Work
Structuring Schedul
CAN
........... :"S.el.ecfir.ig.,.i . /
' sequencing, . ,o """"" .‘
5' 5:312"; ’-. : &sizing : .' Look- ‘-,
:‘ forecasts i I: work we : ‘. ahead 3'
.think can be:
------- insets“: ———q
i "E ..§.|- .ococ.
t'arecaam- 897"?” i
o, ....... ‘ : .9. : 5-.sequencmg: w'" 1...
: Inform .sequencing.&. 5Workable5 n ,gsizing I eeky
g. ation :——>: Sizing work :—V Backlog: : work we f” Work ,
: we think can : i j : know can : \uPlansoy
""" 2.929922-) - be done i l
O........
AA = Asslgnments Anticipated
AMR 3 ASSIQWMMS Made-Ready Resources Production

 

1 PFC = 0105 +WILLS l

 

 

 
   

t; r“: B LI (.1 ‘1 1 £31 81

1: App , l. "
‘,)trlli’.’ll‘.,.lli‘ ’

   

Action to prevent
repetitive errors

A

WILL

 

 

 

 

 

 

 

Figure 2.3‘Last Planner System’ (with additions in red) (Source: Abdelhamid 2008 — adopted from LCl)

In the LPS, front end planning belongs to the project definition and design phases

of projects. One of the products of front end planning is master schedules. Master

schedules primarily demonstrate the feasibility of project completion by target end date.

Those purposes or functions do not require a high level of detail, which most often is

inappropriate because of uncertainty regarding the future. Master schedules are expressed

at the level of milestones, typically by phase. Phase schedules are produced by cross

functional teams using pull techniques close to the scheduled start of the phase. Phase

-26-

schedules feed into lookahead windows, usually 3 to 12 weeks in duration. Lookahead
processes make scheduled tasks ready for assignment; such tasks are placed in Workable
Backlog. Tasks are allowed to maintain their scheduled starts only if the planner is
confident they can be made ready in time. Scheduled tasks are made ready by screening

for constraints, then by assigning make-ready actions to remove those constraints.

The lookahead process generates early warning of problems so there is more time
to resolve them. Weekly work plans are formed by selection of tasks from Workable
Backlog. Every effort is made to make only quality assignments; i.e., those that are well
defined, sound, in the proper sequence, and sized to capacity. The percentage of planned
assignments completed also known as ‘percent plan complete’ (PPC) is tracked and
reasons for non-completions are identified and analyzed to root causes. Action is taken on
root causes to prevent repetition of errors downstream in the project or on future projects

(Ballard 2000).

PPC is the number of completed assignments divided by the total number of
planned assignments, expressed as a percentage. PPC becomes the standard against
which control is exercised at the production unit level, being derivative from an
extremely complex set of directives: project schedules, execution strategies, budget unit
rates, etc. Given quality plans, higher PPC corresponds to doing more of the right work
with given right resources (Ballard 2000). PPC measures the extent to which the front
line team commitment was realized. Analysis of nonconformance can then lead back to

root causes, so improvement can be made in future performance.

-27-

Since its introduction in 1994, the Last Planner system has been evolving to more
improved versions. For example, lookahead planning and measuring ‘percent plan
complete’ (PPC) have been incorporated as important steps in successful completion of
work. PPC indicates the shortcoming of the production planning on site afier a thorough
reasons analysis is conducted. Practitioners on site constantly work towards improving

PPC as they move downstream in the project.

In a study in 1997, Ballard identified that measurement of the PPC of weekly
work plans revealed a chronic and widespread problem of low plan reliability. This
problem was of vital importance because of its adverse impact on labor productivity both
of the production unit that has a low PPC and those downstream production units which
inherit the uncertainty passed onto them.

As depicted in Figure 2.4, improving PPC is critical for project managers because
it increases the time-cost trade off limit by reducing the work flow variation. The benefit
of maximum resource utilization is achieved only if the variation from work processes is
removed, if the work variation is high (PPC is low) in upstream processes, increasing
resource utilization in downstream processes will increase the wait times of crews,

eventually losing the benefit of a time-cost tradeoff exercise.

-28-

Wait
Time

 

Target

 

L-----_----_----------—-----_........

 

 

 

 

50% 65% 80% 100%
Capacity Utilization

Figure 2.4 Improving the trade-off between time and cost by reducing workflow variation
(Source: Ballard et al 2002)

Ballard (2000) proposed lookahead planning as the key to improving PPC, and
consequently the key to reducing project cost and duration. The functions of the
lookahead process (Ballard 2000) are to:

1. Decompose master schedule activities into work packages and operations
2. Develop detailed methods for executing work

3. Shape work flow sequence and rate

4. Match work flow and capacity

5. Maintain a backlog of ready work

6. Update and revise higher level schedules as needed.

These functions are accomplished through various specific processes, including
activity definition, constraints analysis, pulling work from upstream production units, and
matching load and capacity (Ballard 2000). The lookahead process functions in the

following manner; at first, the master and phase schedule activities are exploded into a

-29-

level of detail appropriate for assignment on weekly work plans. Then each assignment is
subjected to constraints analysis, as shown in figure 2.5, to determine what must be done
in order to make it ready to be executed. If the planner is not confident that the

constraints can be removed, the potential assignments are postponed to a later date.

Week 3/25/2002 4/1/2002 4/8/2002 4/15/2002 4/22/2002 4/29/2002
PPC 25% 62% 62% 35% 65%
Tasks 1 8 l3 7 13
leted
Tasks Planned 13 21 20 20

Coordination

Owner Decision
Weather

 

Labor
Figure 2.5 Constraints Analysis sheet (Source: Abdelhamid 2008)

Different types of assignments have different constraints; contract, design,
submittals, materials, prerequisite work, space, equipment, and labor etc. are few
examples of constraints (Ballard 2000). These constraints if not removed cause tasks to
be incomplete. Measuring performance at the Last Planner level not only allows a change
at that level; root causes of poor plan quality or failure can be traced back at any
organization level, process or function (Ballard 2000). The first thing in this process is
identification of reasons why planned work was not done. Some of the reasons that have
contributed to poor PPC are: faulty directives or information provided to the Last
Planner, too much work was planned, failure in coordination of shared resources, change
in priority, e.g., workers reassigned temporarily to a “hot” task, design error or vendor

error discovered, etc (Ballard 2000).

-30-

Several implementation of lookahead planning and reasons analysis have
identified typical reasons for plan failure, such as materials, craft-coordination,
information, changes, emergencies, tools and equipment, design problems, and not
having enough workers on site in the order of decreasing frequency. A case study was
conducted in 1997 by Ballard et al, to track reasons for not achieving 100% PPC. Of the
249 assignments that were not fully completed, 71 were because of late or defective
materials, 42 because prerequisite work was not completed, 37 because of changes in
priorities, 33 because of absenteeism or accident (manpower), 23 because of failure to
accurately estimate the amount of labor time required to execute assignments, and so on
(Refer Figure 2.6). Studies showed that prior to implementation of the LPS lookahead
process, the PPC averaged around 50%. Miles (1998) reported that the overall PPC

improved to 75% with implementation of lookahead process.

I Failure to estimate labor time
I Other Reasons
13% 9% I Late or defective materials

I Prerequisite work not

15% compieted

I Changes in priorities

I Absenteeism orAccident

1 7%

 

Figure 2.6 Pie Chart showing frequencies of various reasons for incomplete work (<100% PPC)

-31-

In a study of a large mechanical contractor’s production control system, Ballard

(1997) reported the performance measurements used by the mechanical contractor for the

improvement of lookahead planning:

Subjective evaluation by project superintendents/managers and consultants.
Assignments Anticipated (Refer Figure 2.3). Measures the extent to which weekly
work plan assignments previously appeared on lookahead schedules.

Assignments Made Ready (Refer Figure 2.3). Measure the extent to which
assignments that appeared on lookahead schedules appeared on weekly work plans
when scheduled.

Change of scheduled dates for specific assignments over time.

-32-

Mitropoulos (2005) built on those findings and developed ‘Make Ready’ metrics to
assess and improve the “make ready” process (Figure 2.7). The first proposed metric

called “Planned Work Ready” (PWR) indicates the portion of the planned activities that

FACTWITIES j ---------------- @¢ .

 

 

 

 

 

 

 

 

 

 

 

 

- ------ ---------- PPC u:
Lookahead 1‘ 23:33:31: WWP: Week-1 '
Plan: Work . work released Work
that should be plerfzrrnedén & workable ‘ I performed
performed l 03032:: backlog ;

 

 

 

 

 

 

 

 

 

 

 

.0000000-

5....

 

 

 

 

 

 

 

 

 

Constraints . Constraints
Constrarnts ,
that should be analysis: Action removed.
removed --—-—> . . .__~ Actron Items
(Action items Items promIsed com leted
. . to be completed p
Identified) (AlC)

 

 

 

 

 

 

 

  

  

  
     

  
   

Als not
promised

Als not
identified

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Eousrrwms
Reasons for Reasons for R f
Constraints Not Promised 00:81:;ng (got
Expected To Be Constraints Not Ant' . ted
Removed Removed “"93
REASONS

 

Figure 2.7 Elements and Metrics of Make Ready Process (Source: Mitropoulos 2005)

-33-

the project team is confident that can be performed in the lookahead horizon. The metric
does not include only work that has all the constraints removed at the time of the forecast,
but also work that is expected to be ready with a high degree of confidence. The main
purpose of the PWR was to establish and track confidence in planned work over a period

of time till the actual work week arrived.

PWR could be represented in two ways (Mitropoulos 2005): a) Percent of
activities ‘Ready’ for each week in the lookahead, and b) Eamable Value. Figure 2.8 and

2.9 illustrates the PWR in its two parts.

 

Planned Work ready
Week 1-5

 

90%
80%
70%
60%
. 50%
40%
30%
20%
10%
0%

— Not Ready

— Ready

No. of Activities
Percent Work Ready

wr- Percentage

 

Weekl WeekZ Week3 Week4 Weeks

Lookahead Week

 

 

 

Figure 2.8 Planned Work Ready for lookahead period (Source: Mitropoulos 2005)

-34-

  

 

 

A ...... . .......................................... -.-.-.-.-._ .......... -

Revised BCWS /,/’"

//
: ---'./-,/-
.- . Z /
a ““““““
’8' E Eamable
'§‘ ' J . Value
0. '.'/'"l'""' "'
9 l/ l
.2 I/ I
E I. :
3 I I
E I l
3 i I
0 I I
Original plan
TIme

Lookahead Horizon
Figure 2.9 Eamable Value (Source: Mitropoulos 2005)

Figure 2.8 illustrates the percent of activities ready for the next 5 weeks in the
lookahead horizon. For example, week 1 lookahead includes 16 activities that are
expected to be ‘ready’ and 4 ‘not ready’, and so on. Figure 2.9 illustrates the ‘eamable
value’ in the lookahead horizon. This would require calculating the eamable man-hours
of the work in the lookahead horizon and comparing with the available labor capacity in
order to decide if the project needs more of less manpower (Mitropoulos 2005).

Mitropoulos (2005) stated that PWR is a forecast metric that would indicate the
momentum of the project and in combination with percent complete it would provide a
better indication of schedule performance. This is a flawed assumption since it has
already been discussed in other literature and contested by researchers that EV does not
indicate time performance. Even the momentum fails if it does not know what obstruction
lies in the path, clearly not perceptible by the EV.

The second make ready metric was developed to assess the accuracy of PWR, this
metric implemented the use of time-time charts and used to compare between forecasts

and actual work for a specified week. Figure 2.10 illustrates a time-time chart. Squares in

-35-

the same column show the difference between expected work and actual work, and how
accurately the organization predicts the upcoming work. Each square i-j indicate that
week ‘I’ is in progress and week ‘j’ is being looked at. For example, FO-l is the forecast
developed on week 0 for week 1 and FO-6 is the forecast developed on week 0 for week
6. the diagonal squares (i-i) show the actual work that is performed on week i (AWi).
Comparisons between forecasts and actual work for a specific week (squares in the same
column) show the difference between expected work and actual work, and how

accurately the organization predicts the upcoming work (Mitropoulos 2005).

....looking at week....j

 

 

 

 

 

 

 

 

 

 

 

0 1 2 3 4 5 6 7 8 9
0 AW-O F0-1 FO-Z F0-3 F0-4 F0-5 F0-6
1 AW-1 F1-2 F1-3 F1-4 F1-5 F1—6 A F1-7
2 AW-2 F2-3 F2-4 F2-5 F2-6 F2-7 F2-8
:3 3 AW-3 F3-4 F3-5 F3-6 F3-7 F3-8 F3-9
6; 4 wks 3 wks 2 wks last . this next 2 wks 3 wks 4 wks 5 wks
8 4 ago ago ago week week week ahead ahead ahead ahead
E 5 AW-5 FS-6 v F5-7 F5-8 F5-9
3 6 AW-6 F6-7 F6-8 F6-9
7 ANN-7 F7-8 F7-9
8 AW-8 F8-9
9 AW-9

 

 

 

 

 

 

 

 

 

 

 

2005)

-36-

Figure 2.10 Time-Time chart showing work forecast and work performed (Source: Mitropoulos

 

The third metric was developed to improve the organization’s ability to remove
constraints and improve the forecasting ability of the organization; this was proposed by
developing Action Items (AIs) indicated on following deltas (refer Figure 2.7):

a) Delta between constraints identified vs. constraints expected to be removed
(promised). These are the constraints that prevent planned work to become ready. This
delta can also be expressed in terms of Ms identified and A15 expected to be completed.
The proposed metric is Al promised/AI identified.

b) Delta between constraints expected to be removed vs. constraints actually
removed. The proposed metric here is Al completed/AI promised.

c) Delta between constraints identified vs. actual constraints found when the
work was released. This is the case where the constraint analysis failed to identify all
constraints during planning. In LPS, this is captured as “planning failures”. The proposed
metric here is Number of new constraints (Als) discovered during execution/Constraints
identified. Further, Mitropoulos proposed the use of reasons analysis to understand why
the identified constraints cannot be removed, why constraints were not removed and why
constraints were not anticipated to help identify bottlenecks e.g., timing of identification,
personnel workload, contractual issues etc. and provide direction on how to increase its
ability to ‘make work ready’. The summary of these metrics are presented in Table 2.3

(Mitropoulos 2005).

-37-

 

Purgose

Metric

Analysis

 

Assess how much planned work
will be ready to perform in the
lookahead period

a- Planned Work Ready (PWR)
% of assignments expected to be
ready as planned

b- Eamable Value (Ballard 97)

3 Value of work expected to be

Reasons for planned work not
expected to be ready.

 

performed

Assess and improve the forecast Actual Work vs. Forecasted 1- Reasons for differences
Work for different lookahead between forecast and actual
horizons. 2- Reasons for difference

% of assignments in both Fi-j and
AWj

between forecasts

 

Evaluate and improve the
organization’s ability to identify
& remove constraints

1- AI promised/AI Needed

2- AI Completed/ AI Promised

3- Constraints not identified
during make-ready process

I- Reasons for constraints that
cannot be removed as needed

2- Reasons for constraints not
removed as promised

 

3- Reasons for constraints not
anticipated

 

Table 2.3 Proposed PWR Metrics (Source: Mitropoulos 2005)

The addition of PWR and other two metrics promise to add great value to the
lookahead planning process by screening constraints at every progressive week and
increasing the ability to deliver work as promised. The third metric, action items,
implements a ‘three level’ constraints tracking method (as discussed earlier) during the
reasons analysis phase to help planners categorize constraints according to their
frequency of occurrence and ease of identification and removal. However, forecasting
schedule performance and measuring the accuracy of the forecast can only add value if
the forecasting tool being used provides an understanding of the nature of relationship
between all the production factors on the outcome of work. This is the focus of this

research.

The first step towards defining such relationship is identification of production
delay factors. Prior research have directed efforts towards identifying these factors in
order to understand the complex nature of construction projects and develop methods to

simplify the planning, execution and control of production tasks. The following section

-38-

 

presents a discussion of previous research that studied production delay factors of

construction projects.

2.5 Production factors of Construction Projects
“Identifying factors that impact construction productivity is not a new effort. There have
been numerous efforts of identifying and classifying the factors that impact construction
productivity, with a few attempting to identify the relative importance of the individual
factors (Dai et al 2009a).” One of the earliest of these related efforts was a United
Nations (1965) study that reported how substantial improvements in labor productivity
can be achieved through repetitive site operations. Borcherding and Oglesby (1974) and
Maloney (1981) examined the effects of craft motivation on construction labor
productivity. Thomas et al. (1989), Thomas and Sanvido (2000), and Horman and
Thomas (2005) examined the impact of material management practices, delivery
methods, and fabricators on productivity.

The loss of productivity as a result of scheduled overtime has also been examined
in the past (Oglesby et al. 1989; Thomas and Raynar I997). Diekmann and Heinze (2001)
examined the influence of support personnel, drawing, equipment, and material buffer
strategies on productivity in the piping and electrical trades. Rojas and Aramvareekul
(2003) conducted a web-based survey to identify the relative importance of 18 factors
affecting labor productivity. Liberda et a. (2003) identified the relative importance of 51
productivity factors categorized under the headings of labor, management, and external
factors by interviewing industry experts. Dai et al. (2009b) in their latest such study have
examined the underlying structure of the factors affecting construction productivity from

the craft worker’s perspective; they identified a total of 83 factors and judged for relative

-39-

importance through a craft worker survey. The study highlighted that management
factors, such as lack of detail planning, inadequate supervision, and lack of information,
were found to account for half of the most critical factors.

The brief review of previous researches on specific productivity factors is not
meant to be exhaustive, since a discussion involving each of those is beyond the scope of
this research document. However, the research acknowledges that the factors addressed
in our research were already known, with some supplemented by the author.

The production factors selected in this research are specific to the Last Planner
production control system discussed in earlier sections of this chapter; the theoretical
relationships that exist among the factors have largely been based on the opinions of a
few selected industry professionals, specializing in Lean Construction, and also drawn
from the previous researches mentioned. These relationships will be presented and

discussed in Chapter 4.

2.5.1 Classification of Production Factors

Multiple studies have presented construction productivity models to explain the
interaction of the productivity factors (Dai et al. 2009b). It is extremely difficult to
distinguish the influence of any single factor since jobsite productivity is simultaneously
influenced by multiple factors (United Nations 1965). Productivity classification schemes
over the years have reflected a change in project characteristics, including an increased
sophistication of project design (Dai et al. 2009b). Herbsman and Ellis (1990) grouped
productivity factors into technological and administrative based on interviews with

industry practitioners; technological factors were primarily related to project design, such

-40-

as specifications, design drawings, and material selection, meanwhile, administrative
factors were defined as being related to management and construction of a project, such
as equipment, labor, and social factors. Thomas and Sakarcan (1994) defined two broad
classification schemes for construction productivity: organizational and executional
continuity. Organizational continuity referred to actual work to be done, such as work
scope and size of the components. Executional continuity consisted of the work
environment and management components, such as weather and work sequencing (Dai et
al. 2009b). Olomolaiye et al. (1998) divided productivity factors into external and
internal factors, representing those factors beyond and within the control of management,
respectively. External factors included the nature of the industry, the construction clients,
weather, the level of economic development, legislation, procurement policies, codes of
practices, etc. Internal factors involved management, technology, labor, unions, and so on
(Dai et al. 2009b). In the study by Rojas et al. (2003), the factors were classified into four
categories, including management systems and strategies (e.g. scheduling), man power
(e.g. experience and motivation), industry environment (e.g. adverse working conditions)
and external conditions (e.g. scope changes).

This research focuses on the productivity factors from the standpoint of the Last
Planner, that is, the focus is on the production factors that affect work flow reliability.
Hence, the production delay factors would be classified viewing their applicability to the
Last Planner System and how they affect the continuity of work. This research proposed
that production delay factors are of two main types. The first type is the factors that
essentially prevent a work (task) from starting up. These factors are identified in the Last

Planner system during the constraint analysis stage and fall into three broad categories

-41-

namely: Pre-requisite Work, Directives, and Resources. These categories represent the
exhaustive list of factors that if present, signify presence of planning issues that prevent
the start of task, for example, coordination issues, regulatory inspections - these are type
of pre requisite work, RFI’s unaddressed, submittals are unapproved - these two are type
of directives, availability of space, labor - these are type of resources (Ballard 1997), and
prevent the planned work from starting.

If all the production factors covered under Prerequisite work, Directives, and
Resources are addressed then it ensures the start of work, however, it does not necessarily
ensure the finish of the work that is started. This is because construction is a dynamic
process and needs constant input of management and control during the entire process.

Experts in Lean production (Toyota Production System) point out that three main
reasons result in workflow issues. These are Muda (Unnecessary work/waste), Mura
(Variation), and Muri (Overburden). Muda is the Japanese term for unnecessary work
(work that does not add value) like multiple handling of materials, doing rework, etc.
Mura stands for variation in production capacity of the crew and it generally occurs
because of factors such as absenteeism or irregular sizing of the work package from
upstream work, slow learning curve, etc. Muri stands for overburden on crew which
generally occurs due to putting overtime regularly, assigning more tasks to crew than
they can handle. Equipment can also be burdened by using it for work that it is not
designed for, etc.

In this research these three sources of workflow issues will be considered as
sources of disrupting the ability of a crew'to finish work started. Hence, two main

categories will be considered as the source of workflow issues. Table 2.4 below indicates

-42-

the two main categories affecting work flow reliability, with major factors affecting each

category listed as well.

 

 

 

 

l‘uctors that Prcwnt the Start ol‘ Work liuctors that l’rcwnt the Finish of Work

a. Pre Requisite Work d. Unnecessary work (Muda)

b. Directives e. Variation in Production Capacity (Mura)
c. Resources f. Overburden (Muri)

 

Table 2.4 Classification of production factors suited to Last Planner System

As previously discussed in this chapter, current production tools do not have the
capacity to measure the causal impact of factors categorized into prerequisite work,
directives, resources, waste, variation and burden. Therefore, this research explored ways
that can help understand the causal relationship between production delay factors and
work flow reliability. This thesis explored a statistical tool called Structural Equation
Modeling (SEM), which is an advanced form of regression analysis and has been
extensively applied in psychological research and has the ability to include both
measured and latent variables in a relationship model. This was found to be a useful
technique because it is not always possible in construction settings to measure all effects
of the different production constraints. SEM provides an opportunity to analyze observed
and unobserved measures of production constraints and their degree of impact on

workflow reliability.

2.6 Structural Equation Modeling: An Introduction
Structural Equation Modeling (SEM) is a technique used for specifying and estimating

models of linear relationships among variables (MacCallum and Austin 2000). Various

-43-

 

theoretical models can be tested in SEM that hypothesize how sets of variables define
constructs and how these constructs are related to each other. For example, a marketing
researcher may hypothesize that consumer trust in a corporation leads to increased
product sales for that corporation. The goal of SEM analysis is to determine the extent to
which the theoretical view is supported by sample data (Schumacker and Lomax 2004).

SEM is primarily used in observational studies and is also applied in experimental
studies. The SEM designs used for observational studies are broken into two categories:
cross-sectional and longitudinal (MacCallum and Austin 2000).

A cross-sectional design is a single-occasion snapshot of a system of variables
and constructs. Its key feature is the concurrent measurement of variables. The use of
SEM in cross-sectional designs is common, with applications to manifest variable, latent
variable, or measurement studies. A notable feature of such models is the specification of
directional influences among variables.

There are two types of longitudinal designs; both involve measurements obtained
from the same individuals on repeated occasions. In one type of longitudinal design
called sequential design, different variables are measured at successive occasions and the
model specifies affects of variables at a given occasion on others variables at later
occasions.

In this type of research, the interest is in the pattern of influences operating over
time among different variables. The sequence and timing of measurements are designed
to allow for these hypothesized effects to operate. In another type of longitudinal design,

a repeated measures design, the same variable or variables are measured at each occasion.

-44-

This is conducted to understand relationships among the repeated measures of the same
variables as well as the pattern of change over time (MacCallum and Austin 2000).

Variables in a model may include both measured variables (MVs) and latent
variables (LVs). LVs are hypothetical constructs that cannot be directly measured
(MacCallum and Austin 2000), or were not measured. LVs are indirectly observed or
measured, and are inferred from a set of variables that we do measure using tests,
surveys, and so on. For example, intelligence is a latent variable that represents a
psychological construct. The observed, measured, or indicator variables are a set of
variables that are used to define or infer the latent variable or construct. For example,
Dow-Jones index is a standard measure of the American corporate economy construct
(Schumacker and Lomax 2004). In SEM, a construct is typically represented by multiple
MVs that serve as indicators of the construct. The general form of SEM consists of two
interrelated sub-models. The first is the measurement model, which relates latent
variables to their observed indicators; the second is the structural model, which estimates
the relationships between the latent variables (Heck and Thomas 2000).

A structural equation model, then, is a hypothesized pattern of directional and
non-directional linear relationships among a set of MVs and LVs. Directional
relationships imply some sort of directional influence of one variable on another. Non-
directional relationships are correlational and imply no direct influence (MacCallum and
Austin 2000).

Variables, whether they are observed or latent, can also be defined as either
independent or dependent variables. An independent variable is a variable that is not

influenced by any other variable in the model but can be correlated to each other; they are

-45-

also called free variables, whereas if a variable is not correlated to any other variable then
it is an independent random variable. On the other hand a dependent variable is
influenced by another variable in the model. Citing from earlier example, the marketing
researcher believes that consumer trust in a corporation (independent latent variable)
leads to increased product sales (dependent latent variable) (Schumacker and Lomax
2004)

Figure 2.11 below presents a sketch of different components of a SE model. In the
figure, F3 and F4 are the latent variables or constructs, V7 to V9 are the measured
variables for F3 and V10 to V12 are the measured variables for F4. The single headed
arrows between LV and their respective MVs indicate the relationship between the

construct and their MVs.

 

 

57—» V7 v10 4— E10

 

 

 

 

 

 

 

 

E8 . V8 F3 F4 V11 I E11

 

 

 

 

 

 

 

E9 ——> V9 V12 <— E12

 

 

 

 

Figure 2. l 1 Sketch showing different components of a structural equation model

The double sided arrow between the two constructs indicates that they are
independent (free) latent variables. A single sided arrow would indicate directional
relationship from one construct to another as shown in Figure 2.12. In Figure 2.12, the

arrow flows from F1 construct to F2 construct, which means that F1 is an independent

-46-

variable or an Exogenous construct, and F2 is a dependent variable or an Endogenous
construct (Hair et al. 2005). It should be noted that En represents error in measurements,

while the Dn term signifies residual values not explained by the measured variables.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E1 ——> V1 02 V4
E2 -——> V2 F1 V5
53 —> V3 \ V6

 

 

 

 

 

 

Figure 2.12 Sketch showing exogenous and endogenous constructs

In the most common form of SEM, the purpose of the model is to account for
variation and covariation of the MVs (MacCallum and Austin 2000). In comparison with
a conventional regression model, wherein, a single dependent observed variable is
predicted or explained by one or more independent observed variables, Path analysis
models (Type 1 SE Models) allow for multiple independent observed variables and
multiple dependent observed variables, and Confirmatory Factor Analysis Models (Type
2 SE Models) consist of observed variables that are hypothesized to measure one or more
latent variables (independent or dependent); for example, diet, exercise, and physiology
are observed measures of the independent latent variable “fitness” (Schumacker and
Lomax 2004). Path models are used when MVs are of primary interest or when multiple
indicators of LVs are not available. Factor analysis, provides for testing models of
relationships between LVs, which are common factors, and MVs, which are indicators of

common factors (MacCallum and Austin 2000).

-47-

2.6.1 Basics of SEM Estimation and Assessment

Statistically, SEM differs from other multivariate techniques in that it is a covariance
structure analysis technique rather than a variance analysis technique. As a result, SEM
focuses on covariation among the variables measured, or the observed sample covariance
matrix (Hair et al. 2005). A covariance matrix is preferred over correlation matrix mainly
because the use of correlations as input can at times lead to errors in standard error
computations. In addition, any time hypotheses concern questions related to the scale or
magnitude of values, then covariances must be used because this information is not
retained using correlations (Hair et al. 2005). Figure 2.13 below shows an example of
structural model with covariances to be calculated (the covariances are shown enclosed in

rectangles on the arrows).

(F8,F5)
(F5, F6)
(F8, F6) (F9, F8)
(F5 F7) F9

(F6 F7) (F8 F7)

.//

Figure 2.13 Sketch showing Covariances that are calculated in a structural model

 

Paths in the model shown in figure 2.13 represent a research question posed to
understand relationships between constructs F5 to F9. For a model like this the Observed
Covariance Matrix is as shown in Table 2.5. The use of a covariance matrix allows the

use of multiple scales for the measuring of the variables. In addition, using the covariance

-43-

does not require standardizing like in regression analysis where the correlation matrix is

implemented for the exact reason.

Var(FS) Cov(F5,F 6) Cov(F 5,F 7) Cov(F5,F 8) Cov(F5,F9)
Cov(F5,F6) Var(F6) Cov(F6,F7) Cov(F6,F8) Cov(F6,F9)

Observed Cov(F 5,F 7) Cov(F 6,F 7) Var(F7) Cov(F7,F8) Cov(F7,F 9)
Covariance =

Cov(F5,F8) Cov(F6,F8) Cov(F7,F8) Var(F8) Cov(F8,F9)

Cov(F5,F9) Cov(F6,F9) Cov(F7,F9) Cov(F8,F9) Var(F9)

 

Table 2.5 Observed Covariance Matrix for model in Figure 2.13 (Adopted - Hair et al 2005)

The unbolded values above the diagonal represent the 10 unique terms that are the
same as those below the diagonal. Given this duplication, covariance matrices are
generally expressed as symmetric matrices, with the unique terms only shown below the
diagonal (Hair et al. 2005). Hence, afier collection of data, it is plugged into the SEM
software and an observed covariance matrix is generated.

The next step in an SEM analysis is to estimate the relationships (arrows between
constructs) using simple bivariate correlations in a system of structural equations. This
process estimates the strength of each relationship portrayed as a straight or curved arrow
in a path diagram (Hair et al. 2005). A description of this procedure is in appendix A.
With estimates for each path, an interpretation can be made of each relationship
represented in the model. The researcher can assess the probability that the estimates are
significant (i.e., not equal to zero) by applying statistical inference tests. The last step in
an SEM analysis involves calculating an estimated covariance matrix and then assessing

the degree of fit to the observed covariance model. The estimated covariance matrix is

-49-

 

derived from the path estimates of the model (calculated in the previous step) by using
the principles of path analysis in reverse. Then by comparing the two matrices SEM can
test a model. Models that produce an estimated covariance matrix that is within the
sampling variation of the observed covariance matrix are generally thought of as good
models and would be said to fit well (Hair et al. 2005).

The difference between the observed and estimated covariance matrices is called
residual and it is the key driver in assessing fit of a SEM model. As compared to a
conventional multiple regression technique where residuals reflect errors in predicting
individual observations, in SEM residuals means how far away an estimated covariance
term is from the observed covariance term for the same two variables. Thus, residuals are
used as the basic indicator of the goodness-of—fit of a theoretical model.

So far we have discussed an introduction and the basic concepts of SEM. Further

literature review will discuss about SEM components and estimation technique in detail.

2.6.2 Components of Structural Equation Modeling

SEM is a technique widely used for confirmatory factor analysis; that is, it is useful for
testing and potentially confirming a theory. A theory is needed to specify .both
measurement and structural models (explained later); modifications to the proposed
relationships, and many other aspects of estimating a model. A theory based approach is
necessary because all relationships must be specified by the researcher before the SEM
model can be estimated; and it involves proposing that a dependence relationship actually
is based on causation. A causal inference involves a hypothesized cause-and-effect

relationship. If we understand the causal sequence between variables, then we can explain

-50-

how some cause determines a given effect (Hair et al. 2005), albeit in an approxiamate,
not exact manner.

In order to test the theory, a research design is required. The research design
involves defining the concepts that the researcher wants to test in a cause-and-effect
relationship in the form of variables. These variables are called latent variables because
they cannot be measured directly but can be represented or measured by one or more
variables (indicators). For example, a person’s attitude towards a product can never be
measured so precisely as to eliminate uncertainty, but by asking various questions we can
assess the many aspects of person’s attitude. In combination, the answers to these
questions give a reasonably accurate measure of the latent construct (attitude) for an
individual (Hair et al. 2005). A measurement model is the model that is drawn to specify
the relationship between such a construct and its indicators is called a measurement
model, such as Fl with Vl-V3 or F2 with V4-V6 (refer Figure 2.12). The measurement
model enables an assessment of construct validity.

The second aspect of a structural equation model is the structural model itself,
which is the dependence or correlational relationship between all the hypothesized latent
constructs (see Figure 2.13 for reference); testing the validity of structural model is the
ultimate goal of SEM analysis (Schumacker & Lomax, 2004).

The main benefits of using the latent constructs instead of directly using measured
variables comes from the criticism of other statistical tools like multiple regression
analysis that it ignores all the potential measurement errors of the observed variables
(886 2005). SEM takes care of this problem by specifying the measurement model which

tests the reliability of measured variables by checking their internal consistency based on

-5]-

how highly interrelated the indicators are. Therefore, in the process, it highlights the
existence of measurement error and accounts for it in the analysis (Hair et al. 2005).

After the latent constructs are established with their measured variables, the first
step is to check for construct validity. Construct validity is the extent to which a set of
measure'items actually reflects the theoretical latent construct those items are designed to
measure. Evidence of construct validity provides confidence that item measures taken
from a sample represent the actual true score that exists in the population. Construct
validity has four components. These are (Hair et al. 2005):

a. Convergent Vglidifl: The items that are indicators of a specific construct should
converge or share a high proportion of variance in common, known as convergent
validity. There are several ways to estimate convergent validity. These are factor
loadings, variance extracted, and reliability.

b. Discriminant Validity: is the extent to which a construct is truly distinct from other
constructs. Thus, high discriminant validity provides evidence that a construct is unique
and captures some phenomena other measures do not.

c. Nomological Validityzfldflce Validity: Nomological validity is tested by examining
whether the correlations among the constructs in a measurement theory make sense. The
matrix of construct correlations can be usefiil in this assessment. Face validity is the
assessment of degree of correspondence of each measured variable and its conceptual
definition. Technically, it is the most important validity test.

After establishing construct validity, the measurement model is developed. Before
we move to that discussion, it is critical to understand the type of latent constructs and

indicators associated with them as they are critical to SEM model design and validation.

-52-

There are two types of latent constructs that exist namely Reflective constructs
and Formative constructs. Reflective constructs are the constructs that are usually viewed
as producing behavior or phenomenon that is captured by their indicators, meaning that
variation in a construct leads to variation in its indicators (Bollen, 1989). In a reflective
model, the latent construct exists (in an absolute sense) independent of the measures
(Coltman et al. 2008). Such indicators are termed reflective because they represent
reflections, or manifestations, of a construct (Roberts and Thatcher, 2009). For example,
behavioral intention to use a system is often operationalized with three reflective
indicators (eg. Davis et al. 1989). Hence, an individual’s change in the latent behavioral
intention construct results in corresponding changes in each manifest indicator of
intention (Roberts and Thatcher, 2009). Figure 2.14 shows a visual example of reflective

construct and indicators.

 

 

Figure 2.14 Sketch showing reflective construct and indicators (Source: Roberts and Thatcher 2009)

-53-

When defining a reflective construct, one views items or indicators as dependent
on a latent variable. For the construct in Figure 2.14, a measured variable can be
represented in an equation form as:

(i)y1= 711111 + 81

Where yi is the ith indicators, n1 is the latent variable that affects it, 81 is the
measurement error for the ith indicator, and 7., is the coefficient giving the expected effect
of m on yi.

On the other hand, Formative constructs are the constructs viewed as being
formed by their indicators (Bagozzi & Fomell 1982). In a formative model, the latent
construct depends on a constructivist, operationalist or instrumentalist interpretation by
the scholar (Coltman et al. 2008). Such constructs are formed or induced by their
measures. Formative constructs are commonly conceived as composites of specific
component variables or dimensions (Edwards and Bagozzi 2000). For example, at
organizational level, knowledge embeddedness may be defined in terms of planning,
analysis, design, and construction knowledge (Purvis et al. 2001). Hence, indicators of
planning, analysis, design, and construction knowledge form the latent variable
knowledge embeddedness. Figure 2.15 shows a visual example of formative construct

and indicators.

-54-

 

 

Figure 2.15 Sketch showing formative construct and indicators (Source: Roberts and Thatcher 2009)

When defining a formative construct, one conceives the indicators as causing the
latent variable. For construct in figure 2.15, a latent variable can be represented in
equation form as:

(ii) n1=ylx1 + ...... +ynxn+§1

Where m and all Xs are deviation scores, the deviation scores do not covary with
the latent variable’s disturbance term (2;), and the disturbance represents all of the
variance in the latent variable not accounted for by its indicators (Bollen and Lennox
1991). In succinct, the formative and reflective indicators have some conceptual
differences and statistical differences that are critical to understand in order to correctly
identify and validate those constructs. Table 2.6 presents the conceptual differences

between formative and reflective indicators.

-55-

 

Concept Formative Indicators Reflective Indicators

 

Causality Formative Indicators are viewed as Constructs are viewed as causes of reflective
causes of constructs (Blalock indicators (Bollen 1989). Reflective indicators
1971). The construct is formed or represent manifestations of a construct.
induced by its measures.
Interchangeable Not interchangeable -"omitting an Interchangeable - the removal of an item does
indicator is omitting a part of the not change the essential nature of the

 

 

construct" (Bollen and Lenox construct. Although every item need not be the
1991). same, researchers need to capture the domain
space of the construct.
Validity Indicators are exogenously Validity of indicators can be assessed through

determined; hence, correlations are the measurement model (Bagozzi et al. 1991).
not explained by the measurement
model (Bollen 1989).

 

 

 

 

 

Table 2.6 Conceptual differences between Formative and Reflective Indicators (Source: Roberts and
Thatcher 2009)

Unlike reflective indicators, the formative indicators are assumed to be
uncorrelated (Barclay et al. 1995). It is important to note that although theoretically
uncorrelated, in practice, formative indicators may actually co-vary. The important thing
to understand is that even if correlated, formative indicators are not interchangeable; in
fact, removing a formative indicator implies removing a theoretically meaningful part of
the construct (Bollen and Lennox I991). The statistical differences between formative
and reflective indicators are listed in the Table 2.7. Due to their statistical properties,
conventional procedures used to assess the validity and reliability of scales composed of
reflective indicators are not appropriate for compostive variables with formative
indicators (Roberts and Thatcher 2009). One way in which researchers can assess the
validity of a formative construct is by including some reflective indicators to estimate a
multiple indicators and multiple causes (MIMIC) model (Diamantopoulos and

Winklhofer, 2001).

-56-

 

 

Concept Formative Indicators Reflective Indicators

Internal Correlations may not be characterized by specific Indicators should be

Consistency patterns (Bollen I984). internally consistent
(Nunnally & Bernstein
1994).

 

Error Variance

Do not have "error" terms; error covariance
represented only in the disturbance term, I.
Disturbances represent all causes of an endogenous
variable that are omitted from the structural model
(Diamanotopoulos 2006)

Represented by error terms.

 

Identification

 

 

Taken in isolation, the measurement model in Figure
2.15 is statistically underidentified (Bollen and Lennox
I991). The model can only be estimated if it is placed
withing a larger model that incorporates
consequences of the latent variable in question
(Bollen 1989). A necessary, but not sufficient,
condition for identifying the disturbance term if that
the latent variable emits at least two paths to other
htent variables measured with reflective indicators
(MacCallum & Browne 1993).

 

A model with three
indicators is identified.

 

Table 2.7 Statistical differences between Formative and Reflective Indicators (Source: Roberts and

Thatcher 2009)

In the MIMIC model, the formative indicators act as direct causes of the latent

variable which is indicated by one or more reflective measures. The inclusion of

reflective measures is necessary for identification purposes (Bollen 1989). According to

Kline (2005), a model is said to be identified when it is theoretically possible to derive a

unique estimate of each parameter. To identify a model with formative constructs,

scholars suggest (I) placing formative construct within a larger model and (2) specifying

at least two paths from the formative construct to reflective constructs or indicators

(MacCallum & Browne 1993). If a model lacks more than one path to a reflective

construct, the residual variance of the formative construct will be under—identified and

-57-

 

must be fixed at zero. A MIMIC model meets all these conditions, as shown in Figure

2.16 below.
81 82
A1 A2

 

 

Figure 2.16 Sketch showing a MIMIC model (Source: Roberts and Thatcher 2009)

-53-

The construct validity for a formative/MIMIC model can be done by investigating
the significance (should not be equal to zero) of the parameter estimates for each
formative indicator (Bollen 1989). However, retaining non-significant indicators that
contribute to the content domain of a formative construct is considered to be an

acceptable practice (Jarvis et al 2003).

2.6.2.1 Development of Measurement Model
The first step after defining the constructs and undertaking the construct validity tests is
development of the overall measurement model. The individual constructs brought
together to develop the model must be checked for unidimensionality. Unidimensional
mean that a set of measured variables have only one set of underlying construct.
Allowing a single measured variable to be caused by more than one construct impacts
unidimensionality; it is important because the existence of significant cross-loadings is
evidence of a lack of construct validity. In addition, when a measurement model also
hypothesizes no covariance between or within construct error covariances, meaning they
are all fixed at zero, the measurement model is said to be congeneric. Congeneric
measurement models are considered to be sufficiently constrained to represent good
measurement properties, and are hypothesized to have construct validity and consistent
with good measurement practices (Hair et al 2005).

Another critical aspect of developing a measurement model is determining the
number of measured variables in a construct. The number of variables is critical for
model identification purposes. The identification issue deals with whether enough

information exists to identify a solution to a set of structural equations. The sample

-59-

covariance matrix provides this information; one parameter can be estimated for each
unique variance and covariance amoing p measured items, calculated as I/2[p(p + 1)]. One
degree of freedom is then lost or used for each parameter estimated (k). There are three
levels of identification, these are (Hair et al 2005):

a. Under-Identified: An under-identified model is one with more parameters to be
estimated than there are item variance and covariances (i.e., there are negative
degrees of freedom). For example, a measurement model with only two measured
items and a single construct is under-identified. The covariance matrix would be 2x2,
consisting of one unique covariance and 2 error variances. But, the measurement
model of this construct would require that two factor loadings and two error variances
be estimated (four parameters). Thus, a unique solution cannot be found because there
are more parameters to be estimated than unique values in the covariance matrix.

b. Just—identified: With the same logic as discussed above, a three-item indicator
is just-identified, meaning it includes just enough degrees of freedom (0) to estimate
all free parameters. Since all of the information is used, confirmatory factor analysis
will reproduce the sample covariance matrix identically. In SEM terminology, a
model with 0 degrees of freedom is referred to as saturated. The resulting x2
goodness-of-fit statistic also is 0, therefore, experts say that a just-identified model do
not test a theory.

c. Over-identified: These models have more unique covariance and variance terms
than parameters to be estimated. Thus, for any given measurement model a solution
can be found with positive degrees of freedom and a corresponding 2 goodness-of-fit

value. At minimum, a four-item, unidimensional measurement model produces an

-60-

overidentified model for which a fit value can be computed (degrees of freedom 10-8
= 2). Increasing the number of measured items only strengthens this result.

After checking for model identification, measurement scales are set for the
measured variables. All the items indicating a construct need not be of the same scale
type, nor do different scale values need to be normalized prior to using SEM, however,
combining scales with different ranges can sometimes require longer computational time,
therefore, using same scale type is preferred approach (Hair et al 2005).

The next step after developing the measurement model is to specify the
measurement model. This is done by entering the data into the SEM software and running
the process. However, selecting a proper sample size is an area of big debate in SEM
literature. In the following section, we will discuss what is the minimum sample size

suggested by experts for running an SEM successfully.

2.6.2.2 Sample Size in SEM
SEM in general requires a larger sample relative to other multivariate approaches. Expert
opinions regarding minimum sample sizes have varied; they recommend considering the
following aspects to calculate the required sample size for the SEM. These are:

1. Multivariate distribution of the data.

2. Estimation technique.

3. Model complexity.

4. Amount of missing data.

5. Amount of average error variance among the reflective indicators.

-6i-

Some experts (Kline 2005) have suggested an absolute minimum sample size to
have any hope of getting good results would be N = 100. But that would be a small
sample and runs considerable risk of producing invalid results with Maximum Likelihood
Estimation technique (MLE), while 100-200 would be a medium sample size, and .
over 200 would be generally considered a large sample. In cases with sufficiently large
number of parameters to be estimated, Kline (2005) suggested that an ideal case would be
20 cases/parameter, but that 10 cases/parameter is more realistic. Raykov & Marcoulides
(2006) also suggest 10/cases/parameter as a guideline. Kline (2005) suggests that
anything less than 5 cases/parameter would probably not yield trustworthy results.
MacCallum et al (1996) provided a method for determining the minimum sample size
necessary to achieve a given level of power for tests of model fit. Recent work indicated
that minimum sample size necessary to accurately recover population factor loadings is
highly dependent on characteristics such as communality level of the MVs (MacCallum
and Austin 2000), that is, models containing multiple constructs with communalities less
than 0.5 require larger sizes for convergence and model stability (Hair et al 2005). A final
decision on sample size can only be reached after the measurement model is identified.

After the data is collected of the desired sample size, the measurement model is
specified in order to test the measurement theory. Since a latent factor is unobserved, it
has no metric scale, meaning no range of values, thus, it is provided in one of two ways
(Hair et a1 2005):

l. A scale can be set by fixing one of the factor loadings and setting its value (1 is

a good value).

-62-

2. The construct variance can be set to a value. Again, 1 makes a good value.
Using a value of I, for example, results in a correlation matrix of the relationships
between constructs.

Once, the model is specified it is revisited for issues related to identification. The
problems in identification are detected by the following symptoms:

a. Very large standard errors for one or more coefficients.

b. An inability of the program to invert the information matrix (no solution can be
found).

c. Wildly unreasonable or impossible estimates such as negative error variances,
or very large parameter estimates, including factor loading and correlations among
the constructs (absolute value of 1.0).

(1. Models that result in differing parameter estimates based on the use of different
starting values.

In SEM, however, model estimates should be comparable given any set of
reasonable starting values. Thus, to check for identification issues, a confirmatory factor
analysis model is first estimated and parameter estimates are obtained. Next, the
coefficients are fixed to their estimated value and model is rerun. If the overall fit of the
model varies markedly, then identification problems are indicated.

These problems occur mostly due to (I) the number of variables are misspecified,
e.g. there are a total of 12 variables but researcher specified number of variables as only
11. In this scenario, the covariance matrix is not identified. (2) When a variable is
mistakenly specified twice, in this case the covariance matrix will be non-positive

definite and no unique solution would be found (Hair et al 2005).

-63-

Once the model is specified it is estimated with the help of empirical data, the
measurement model is tested for validity by using goodness-offit indices. Goodness-of-
fit (GOF) indicates how well the specified model reproduces the covariance matrix
among the indicator items. The most widely used GOF measure is the chi-square ()6)
index. The difference in the covariance matrices (S-Zk), where S is the observed
covariance matrix and 2k is the estimated covariance matrix, is the key value in assessing
the GOF of any SEM model. This difference would be zero if the researcher’s theory
were perfect. To the extent that perfect fit is not the case, the chi-square value increases;
because the critical values of the chi-square distribution are known, the probability (p-
value) that any observed sample and estimated covariance matrices are actually equal can
be found. In contrast with other regression techniques, for the chi-square GOF test in
SEM, the smaller the p-value, the greater the chance that observed sample and estimated
covariance matrix are not equal. Thus, in SEM a larger p-value (meaning statistically
insignificant) is desired, the best scenario being a small chi-square value and a
corresponding large p-value that is indicative of statistically insignificant difference
between the matrices (Hair et al 2005).

Some other popular model fit statistics used in SEM technique are Goodness-of-
Fit index(GOF), Root Means Square Residual (RMSR) and Standardized Root Mean
Residual (SRMR), Root Mean Square Error of Approximation (RMSEA), etc. The
indices mentioned and discussed are all absolute fit indices. That is they measure how
well the model specified by the researcher reproduces the observed data, and they do not
compare the GOF of a specified model to any other model; this is done by parsimony fit

indices. A popular parsimony fit index used is Parsimony Normed Fit Index (PNFI),

-64-

discussion of which is excluded as it is beyond the scope of this research. Another type of
indices used are incremental fit indices, they assess how well a specified model fits
relative to some alternative baseline model. A popular incremental fit index used is called
Normed Fit Index (NFI) (Hair et al 2005).

Like the reflective models, model fit is assessed for MIMIC model through
standard measures of fit mentioned and discussed above such as model chi-square, CFI,
GFI, NFI, RMSEA or SRMR (Roberts and Thatcher 2009). The best choice of indices to
be used for measurement model validity depends upon the model under consideration and
were discussed in chapter 4 in a demonstration. The guidelines for establishing
acceptable and unacceptable fit were also discussed in chapter 4, as they are
recommendations based on number of variables and sample size selected (Hair et al

2005)

2.6.2.3 Development of Structural Model

After testing the measurement theory, the final stage in SEM involves specification and
validation of the structural theory that is originally proposed by the researcher. The first
step in the process involves specifying the model using path diagram. The structural
theory is represented by specifying set of relationships between the constructs. In the
structural model, two-headed arrows from the measurement model indicating correlations
between all the constructs are replaced with a smaller number of one-headed arrows
(directional relationships). In essence, the structural theory is created by constraining the
covariance matrix using the set of free and fixed parameters representing hypothesized

relationships (Hair et al 2005).

-65-

The issues of sample size and identification are easier to satisfy in a structural
model. If both the conditions are satisfied for the measurement model, they are likely to
be satisfied for structural model because the structural model is nested within the
measurement model and is more parsimonious as it contains fewer estimated paths. The
structural model is identified as long as there are no interaction terms included, the
sample size is adequate (satisfies the measurement model), and a minimum of three
measured variables per construct is used (Hair et al 2005).

After the model specification, special attention is needed to the design of
construct loadings in the model. The measurement portion of the structural model
consists of the loading estimates for the measured items, error variances and the
correlation estimates between exogenous constructs. There are two approaches to go
ahead with estimation of the structural model; first approach suggests that the estimates
that are obtained from the measurement model should be fixed in the structural model,
because otherwise the loadings will change due to the changes imposed during the
transformation from measurement to structural model. The purpose behind fixing the
estimates is to prevent the condition of interpretational confounding which essentially
means that the measurement estimates for a construct are significantly affected by
relationships outside of the specific measures used for the constructs. By fixing the
parameters, estimating the structural model becomes easier due to lesser number of
values to be estimated, however, this approach may lead to change in fit between the
measurement and structural model due to problems with the fixed measures instead of the
structural theory. The second approach allows estimating coefficients for all the loadings

and error variances along with structural model coefficients. This approach helps in

-66-

revealing the interpretational confounding by comparing the measurement model loading
estimates with those obtained from the structural model estimates (Hair et al 2005).

The last step in the SEM analysis is to assess the structural model validity. Since,
the estimated covariance matrix is computed based on restrictions (pattern of free and
fixed parameter estimates) the estimated covariance matrix developed from the structural
model will include more restrictions because more paths are set to 0 as compared to
measurement model. As a result, the structural model cannot have a lower chi-square
value than that obtained in measurement model. The validity of the model is tested in two
steps:

a. The first step is to assess overall structural model fit, as the measurement model
was assessed. The recommended technique is to assess the fit by using one absolute
index, one incremental index and a model chi-square test at minimum. Experts
recommend that one of the indices should be badness-of-fit index (RMSR or SRMR).
The guidelines for acceptance of fit are same as the ones used for measurement model
(Hair et al 2005).

b. The second step is to compare the measurement model fit and structural model
fit. A structural model fit cannot fit any better (have a lower chi-square) than the
overall measurement model; therefore, it can be said that structural theory lacks
validity if the structural model fit is substantially worse than the measurement model
fit (Hair et al 2005).

Reaching a good fit is not sufficient to prove that the structural theory is right.

The researcher is required to examine the individual parameter estimates against

corresponding predictions or paths, each representing a specific hypothesis. Once, the

-67-

researcher confirms the estimates with the theory (in the predicted direction) and check
for statistical significance of the fit, the structural model is said to be completely

validated (Hair et al 2005).

2.7 Need and Application of SEM in Construction

As briefly mentioned earlier, SEM has been extensively used in behavioral and
psychological research. However, it has been only recently used in the construction
industry. “The limited research studies reported in the management literature that have
addressed the problem of improving the project planning process have focused mainly on
examining the individual impacts of a variety of influence factors on the effectiveness of
project planning efforts (Islam et al. 2005).” Laufer and Cohenca (1990) examined the
effect of eight situational variables on the efforts invested in construction project
planning. The study showed how these situational variables individually influence project
planning. Faniran et al. (1994, 1998) evaluated the influence of situational factors in
project environments and organizational characteristics of performing organizations on
project planning efforts and project planning effectiveness. Dvir et al. (2002) evaluated
the relationship between three measures of planning efforts (development of functional
requirements; development of technical specifications; and implementation of project
management processes and procedures) and four measures of project success (meeting
planning goals; end-user benefits; contractor benefits; and overall project success). The
study supported previous research findings which found a significant positive relationship

between the amount of efforts invested in the project definition and technical

-68-

specifications functions of the project planning process on one hand, and project success
on the other (Islam et al. 2005).

“The studies discussed and other related studies have contributed to understanding
how the project planning process interacts with its environment, the potential applications
of the findings to the development of strategies for improving project/production
planning effectiveness is limited by inherent deficiencies in the methodologies applied”
(Islam et al 2005). Firstly, the methodologies used in the previous studies have focused
mainly on measuring directly-observable influence variables, and assessing the effect of
these variables on the project/production planning process. However, the methodologies
did not consider the interrelationships that exist within the variable sets of the influence
factors in order to understand how individual influence factors work together in
influencing planning effectiveness (Islam et al. 2005). SEM provides an opportunity to
measure collective influence of factors by arranging all the factors in a measurement
model and estimating the influences together. Secondly, a major limitation of the
previous studies is that the findings were drawn mainly from a multiple regression
analysis of qualitative data that had been converted into abstract numerical scales. The
conversion of qualitative data to quantitative data leads to a large potential for error in
measurement of a variable. Regression analysis cannot accurately account for the errors
in measurement that occur as a result of this transformation. However, SEM also
accounts for measurement errors, thus producing more accurate representations (Islam et
al. 2005).

The SEM approach has been increasingly applied in the field of construction

engineering and management to explore complex phenomena and dynamic relationships

-69-

(Kim et al. 2009). Mohamed (2003) used SEM to model the joint venture performance of
overseas construction projects. Molenaar et al. (2000) suggested using SEM to predict the
possibility of disputes at the early stage of a project. Islam et al. (2005) modeled the
impact of project conditions on planning effectiveness, through SEM application. Wang
et al. (2005) also proposed an SEM-based prediction model to ensure that partnering can
be successfully implemented to realize probable benefits.

Kim et al. (2009) conducted the latest such research using SEM to model the
relationship among various factors that affect the project success of international
construction projects. The study adopted 64 performance influencing variables on project
performance, initially classified into five categories but later conceptualized into 14 latent
variables with the help of literature review and expert interviews. The study further
compared the developed SEM model with a multiple regression model and an Artificial
Neural Network (ANN) based prediction model using the same influence variables and
concluded that the SEM model had a capacity to predict the project performance with
moderately higher accuracy. While the overall accuracy for multiple regression model
was estimated at 86.3% and average deviance of profit level was 0.82, the figures for
ANN model were 88.8% and 0.67. The SEM model had an accuracy of 90.7% and only
0.56 of average deviance, thus clearly showing that SEM is more accurate and powerful
in recognizing the complex structures of variables and offering some insights into
underlying cause—and-effect relationships (Kim et al. 2009).

Kim et al. (2009) further discussed the advantages of SEM over other statistical

tools for prediction purposes. These were:

-70-

a. The regression method does not identify all the relations necessary to reflect
realistic situations and cannot cope with a complex problem such as hierarchical
structures of dependencies between each factor. In contrast, the SEM can be used
where the final outcome is best represented as a sequence and relation of interrelated
variables. The SEM is also used to recast a complex problem into several smaller
related path diagrams.

b. The SEM can measure direct as well as indirect effects among the various
latent and observed variables. It also enables representing these multi-layered causal
or correlational relationships and their degree of impacts toward the output variables
by providing the structural coefficients. Thus, the SEM allows for intuitive
apprehending of the interrelationships of the variables that are not visible in both
regression and ANN model.

c. The SEM performs better in supporting the process of strategic decision
making. The firm can choose to negotiate a favorable resolution process based on the
SEM feedback; through this feedback system, the firm can choose well-fitted
strategies designed to improve the firm’s capacity to perform or to improve a
project’s particular conditions.

The SEM applications briefly mentioned and discussed above were all designed
using a reflective model approach. The importance of SEM analysis are well established
in the review, however, it is important to discuss the current trends emerging in
approaching the SEM technique and inform the research to progress in the right direction

using that information.

-71-

2.7.1 Current Trends in Application of SEM Technique

The use of SEM analysis technique in management literature is extensive
(Diamantopoulos et al. 2008; Coltman et al. 2008; Ruiz et al. 2008). Traditionally,
management scholars identify structural relationships among latent, unobserved
constructs by statistically relating covariation between the latent constructs and the
observed variables or indicators of these latent constructs; consequentially, they assume
that this relationship between construct and indicator is reflective. With reflective
measurement models, causality flows from the latent construct to the indicator (Coltman
et al. 2008). However, not all latent constructs are entities that are measurable with a set
of positively correlated items (Edwards et al. 2000). Several researchers (Ruiz et al. 2008;
Roberts and Thatcher 2009; Coltman et al. 2008) have proposed and recommended
formative model assessment over existing reflective models in their areas of research, i.e.
business studies, and information systems, by showing results indicating a better fit and
prediction capacity.

Ruiz et al. (2008) have strongly argued in their study of relationship between
customer value and service value that constructs like service value cannot be formulated
as reflective constructs, because, with reflective measures, all components are expected to
covary with one another. However, the benefit component of service value may not
correlate with sacrifice component, for example, a bank may reduce a customer’s
perceived sacrifice by opening a neighborhood branch, which saVes the customer some
time while using its services, but it doesn’t necessarily mean that the benefits of the bank
for the customer has changed (allowing more withdrawals per day than earlier. Thus,

models using reflective components may misspecify the customer value construct, which

-72-

can cause biased estimates of the structural relationships between constructs and
undermine the validity of the statistical conclusions (Mackenzie et al. 2005). Similar
model misspecification examples and arguments can be found in other research (Roberts
and Thatcher 2009; Diamantopoulos et al. 2001). In the wake of this overarching
misspecification problem, several researchers have recommended steps to investigate
whether a construct should be reflective or formative.

Coltman et al. (2008) proposed a set of theoretical considerations that should be
considered in deciding whether the measurement model should be formative or reflective.
These are:

l. The Nature of the construct: In a reflective model, the latent construct exists
independent of measures, for example, measures of a person’s attitude or personality
do not dictate or indicate whether attitude is there or not — we know it is there but
can’t directly measure it. In contrast, in a formative model, the latent construct
depends on a constructivist, operationalist or instrumentalist interpretation by the
scholar. For example, the human development index (HDI) does not exist as an
independent entity. Instead, it is a composite measure of human development that
includes: health, education and income. Any change in one or more of these
components is likely to cause a change in a country’s HDI score, not the other way
round.

2. Direction of Causality: The second key consideration in deciding a reflective or
formative construct is the direction of causality between the construct and indicators.
For example, since construction production planning presents a proactive scenario,

variation in productivity does not cause or reflect absenteeism in workforce; instead,

-73-

absenteeism in workforce causes variation in productivity. Therefore, it is crucial for
the construction management researchers to design the construct carefully (an aspect
of SEM not addressed in previous researches).

3. Characteristics of Indicators: In a reflective model, change in the latent
variable must precede variation in the indicator(s). Thus, the indicators all share a
common theme and are interchangeable. This interchangeability enables researchers
to measure the construct by sampling a few relevant indicators underlying the domain
of the construct. Inclusion or exclusion of one or more indicators from the domain
does not alter the content validity of the construct. However, in formative models,
since the indicators define the construct, the domain of the construct is sensitive to the
number and types of indicators the researcher selects. Adding or removing an
indicator can change the conceptual domain of the construct significantly. For
example in construction production planning, measuring the availability of resources
(latent construct) by indicators such as availability of tools and availability of labor,
represent a conceptually more accurate Resource construct. Replacing availability of
tools with labor cannot justify the construct, because these two are uncorrelated items
and conceptually alter the definition of availability of resources.

The discussion above about the nature of the latent constructs and measurement
model with examples in management literature recommending taking the formative
indicator approach were used in this research to define the nature of the model and design
the method for exploring the relationship between work flow reliability and production

delay factors.

-74-

2.8 Chapter Summary

This chapter provided an overview of current production management practices
discussing the conventional and lean production management tools. Existing researches
on production factors affecting construction projects were discussed as part of the
literature analysis in fulfillment of research goal and objectives. The Structural Equation
Modeling analysis technique was extensively reviewed and its nuances were highlighted.
The need and existing applications of SEM analysis technique in construction
management were discussed. Overall, the literature analysis brought forth essential
production delay factors, structural equation modeling techniques, and the current
direction in SEM analysis which are all relevant in developing the intended method in
this research to explore the relationship between work flow reliability and production

delay factors.

-75-

CHAPTER-3

RESEARCH METHOD

-76-

3.1 Introduction
The goal of the research is to improve the understanding of factors affecting workflow
reliability of construction projects at the production level. To approach this goal, the
research focused on developing a method to investigate the causal relationship between
production delay factors and their impact on workflow reliability. To accomplish this, the
following objectives were proposed:
1. Study production control tools implemented at the site level and document the
production delay factors encountered on construction site.
2. Develop a method to study the relationship between production delay factors and
work flow reliability.

Figure 3.1 shows the process adopted for the research, which is comprised of four
phases, which were followed to address objectives one and two. The first step was crucial
in achieving the first objective and the remaining three steps helped in achieving the
second objective.

The research began with a comprehensive literature review to identify the existing
methods used in evaluating the performance of production management as implemented
on construction projects, and causes of variation in work flow, and Structural Equation
Modeling analysis technique. In the next step, the research identified the scope of
development of the framework. Next, the development of a framework to design models
capable of understanding the relationship between production factors and reliable work
flow (construction site performance) took place. The framework steps were explained

with the help of a demonstrated example.

-77-

 

Literature Analysis

 

 

 

 

 

 

 

 

 

 

 

 

 

Phase 1
Establish Scope for Develop and
Development Demonstrate Framework
Phase 2 Phase 3

 

 

Discuss Conclusions,
future areas of research

 

 

 

Phase 4

Figure 3.1 Research Methodology in 4 phases

The development of the framework to construct SEM model suited to
construction industry research is expected to lead to the understanding of the causal
relation between the production delay factors leading to unreliable workflow. Given the
qualitative nature of most of the factors affecting production performance and workflow
reliability, the techniques used in Structural Equation Modeling (SEM) will be
instrumental in developing the proposed statistical procedure.

Figure 3.2 shows the detailed research methodology adopted where the steps have
been broken down to detailed task levels to fiirther explain the process. The outcome of
each step is a deliverable and becomes the input for the next step.

In Phase-l as shown in Figure 3.2, the literature review is broadly divided into
three categories: Lean and Conventional production management and control techniques,
literature on construction productivity factors, and structural equation modeling and its

applications. The literature on project and production management helped to understand

-73-

the current shortcomings of macro-level project controls as well as identifying the typical
production delay factors encountered in practice. As mentioned earlier, Structural
Equation Modeling (SEM) is adopted in this research as a statistical tool to explore the
possibility of studying the production factors and work flow reliability in a causal
fashion. The background study on SEM helped in identifying the steps and needed to
prepare the framework to build a causal relationship model.

Phase-2 in the research involved establishing the scope of framework
development for SEM model specific to application in construction management
research. In this step, deliverables expected at the end of the fiamework development are
outlined. In Phase-3, the fi'amework for designing causal equation models was developed
and explained using a demonstrated example in context of construction production
planning. This step also produced a survey instrument that will help Last Planners record
the level of existence of production factors and study their impact on reliable work flow.
In addition, strategies to aid development of the SEM model were also established (only
valid for formative models). In Phase-4, discussion on conclusions from the framework

development process, benefits of research, and future areas of research were presented.

-79-

3268562 :88QO 338 mm 83mm

 

ISHCIIARKI HDHVHSERI

HDGH'IMONX GNU-SIXTH

 

 

note—P533
8 953% Sam .3.“ gee-$052.,—
ue «define—96n— ue 2.3m amass-mm

........... L

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. Boa
emfiofl Epoch 8% co metaboo #53 wfimaeafi E 836358
2mm Co 8:553 @383 coco-c cocoa-van @383 «Become-HE cocoa-ca
.6 muons-E £662
3038:.“an m: EB 88on 5:58:35 moscmfiooh 65:00
maze-662 8356mm 123035 coco—5980 E ""3an EB Sofie «:32 832605

3:232:80 use :34
a w E .-

_ Borrow— ouaaaoxfl _

 

 

 

 

 

 

 

 

 

 

 

Z HSVHJ

 

 

I EISVHJ

 

 

-30-

 

NOILIIHIIILNOD HDHVEISEIH

5:88 3 use..—

 

 

aafi «flea...

 

 

 

:85qu .«e mace-«e. 9.3.5 @553
can memo—Sm anew-@2050 3:85

 

1% .....................

 

feta—:9:— 2:
meme: .23:— Emw we «sonic—gov
ea 3 mom-.35 eases—

 

 

 

 

 

: 6382569
9.83.. 559.59..—
=e 55:58.: 9.3an
3 «haeeeumoaO
Siam augmen—

 

 

3. 6382568
5.5 ace-56%:
\B Siam use: vantage—on—
e...“ 3.5:. Sam we Eon—883‘
a. ewe-won— ..8 xueBoEau-m Ame—95¢

 

 

 

 

 

 

mum—gm 20“...—

I? flSVHd

E C'ISVHJ

 

 

 

-8]-

3.2 Existing Literature, Scope of Development and Research Contribution

The analysis of existing literature in Phase-l provided guidelines for conducting the
research. The outcome of literature analysis provided clear directions for establishing the
extent to which the intended framework is to be developed. The current production
planning and control tools, and literature on SEM applications provide the main research
drivers of the research. The applicability and utility of the Structural Equation Modeling
will act as a driver for establishing scope of the intended framework, in Phase-2.

The main outcomes of the research are delivered from Phase-3, where the
research developed (1) the survey instrument to capture the extent of presence of
production delay factors, and (2) the framework to design causal relationship equation
using SEM in the production planning and control context. This phase includes method of
recording the production delay factors, sorting them into categories required to fit the
causality model, method of populating the model with the recorded data, etc. The
framework was demonstrated along each step using the developed survey instrument with
the help of randomly simulated hypothetical data in this phase of research. In addition,
strategies to support the development of SE model were also developed.

Phase-4 discussed how the framework development benefits the studies on
establishing cause-and-effect relationships between different elements of construction

process.

3.2.] Phase-1 Literature Analysis

Objective-I Study production control tools implemented at the site level and document

the production delay factors encountered on construction site.

-32-

Objective 1 is fulfilled by performing following steps:

a. Conduct literature review on production planning and control tools,
productivity measurement studies and identify factors causing work flow
variability.

The following section discusses the contribution to the research of literature analysis of
conventional production management and control techniques, Lean production

management and control techniques and Structural Equation Modeling.

3.2.1.1 Conventional Production Management Techniques

The literature analysis of conventional production management techniques will identify
attributes of production planning and control tools that are implemented to measure,
report and remediate the actions on the job site in order to achieve best results by most
production planners in the industry. The purpose of reviewing these tools is to understand
the criteria they use to measure progress, methods of measuring the progress, stages at
which the progress is measured and the inferences that are drawn out of those
measurements.

The analysis uncovered how the inefficiencies of the tools were addressed in
previous research (Singh 2007). For instance, the main limitation of the current
production control tools like CPM and EVM is that both tools are sensitive to
disturbances in overall project progress and give inferences that only look at a short term
solution without understanding the depth of the problem. These tools do not aid in
preempting problems but merely act to subdue the recurring problems by reflecting

possible solutions to patch the project performance. Both tools lack the ability to shield

-33-

the production tasks from the production constraints and variations because there is no
consideration of the interdependence between activities.

The LOB method of scheduling, another widely used production control tool
mainly in Europe, has a slight advantage over the CPM method, because LOB method
primarily calculates the schedule based on production rhythms (discussed earlier) and
consider space constraint as an important input before planning out the project, which
helps in preventing bottlenecks on project sites due to overcrowding of crew and shield
the production task from invariability of work flow due to space availability. However,
LOB is not suitable for phases of projects that have non-repetitive elements, such as
footings.

The literature analysis captured the problems and studied the progression of such

tools and the endeavors by different researchers to overcome those problems.

3.2.1.2 Lean Construction Management Techniques

The literature of Lean Construction was be analyzed to understand the development of a
new and integrated project management method called the Last Planner system and its
advantages over the traditional tools that were discussed in the previous section. For
instance, the LPS implements operational shielding of production tasks from various
constraints such as space constraint, quality constraint, material constraint etc. The
literature discusses how the operational shielding of tasks has resulted in improved
production performance but at the same time highlights a general lack of understanding
of the complexities of a combined effect of production constraints on the final

performance. A detailed study of all stages of implementation of LPS helped identify the

-84-

different factors that plague the project and production performance and would facilitate
ideas for improved production control.

The research on subsequent developments on LPS helped to identify the direction
of progress in research and help pick the attributes of the production planning system that
require further research and development. For instance, the PWR metric helped in
increasing the confidence of production planners in their plan over the period of six
weeks leading to the production week, but is unable to build confidence in the outcome of

the production plan and develop strategies to handle inefficiencies and uncertainties.

3.2.1.3 Structural Equation Modeling

The literature analysis of SEM was performed to understand the applicability of the tool
to this research. The review identified the process of using SEM for any study in general,
and then modified it to suit this research. The review discussed different types of SEM
designs and identified the appropriate design for this research, for instance, whether a
cross-sectional design or a longitudinal design is best suited for this research, or a
formative model is preferred over reflective model. The methodology of conducting an
SEM research was documented and each step will be discussed to identify its ingredients,
the process will help in comparing the nature of research data required to design and
populate a SE model and the data available on construction projects, for instance, which
production factors can be classified into measured variables or latent variables; and how
to collect information of these factors. The sample size of the data required to get a good

fit for a causality model depends on the type of design used by the model.

-85-

Given the exploratory nature of this research, the literature review allowed
defining the scope of application of SEM, such as whether unique measurement methods
of production factors should be adopted to build a universal relational model between
production delay factors and work flow reliability. Figure 3.3 presents the different steps
involved in designing a structural equation model. A discussion on each critical step
helped establish the selection criterion for data and develop methods to convert data into

usable formats.

 

Model Specification

 

 

 

v
Model Identification

 

 

 

 

 

 

 

 

 

Model Estimation

 

 

 

 

 

Model Testing

 

 

v
Model Modification

 

 

 

 

Figure 3.3 Five basic steps of designing a Structural
Equation Model (Source: Schumacker and Lomax 2004)

The steps listed above form the core process of pursuing a relational equation
using SEM. The discussion of these steps and the sub steps will contribute to
development of the framework. The literature analysis will also reveal the software of

choice for processing data of the structural equation model.

-86-

3.2.2 Phase-2 Establish Scope of Development
Objective-2 Develop a method to study the relationship between production delay
factors and work flow reliability.
Objective 2 is fulfilled by performing following steps:

b. Develop a Survey Instrument to measure the existence of identified

production factors.

c. Based on the literature review, develop a framework to design a causality

model between various production factors and work flow reliability.

After the literature analysis exposed the current limitations of all three systems
discussed, the essential attributes of the three areas were studied to establish the
boundaries of exploration that will be reached during this research.

The first research question considered the extent with which the Last Planner
system can help forecast construction site performance. Using the knowledge gained
from the literature analysis, this step identified the next logical steps required to
extend/create the forecasting capability of the Last Planner system. A discussion on the
extent to which the Last Planner system is SEM-ready will help in deciding the desirable

outcome of the research.

3.2.3 Phase-3 Develop Framework for designing Causality Equation

After reviewing the literature and establishing the scope of SEM development, a
framework was developed to outline the sequence of activities that are to be adopted by
production planners when developing a relational equation using SEM. A detailed

discussion of the framework will illustrate how to measure production performance,

-37-

collect and analyze data, developing the SEM model, estimating the model, adjusting the
model, reporting and confirming the theory. The process map will also include steps that
describe the transformation of project data to useful research data. The framework is a
first step towards the development of prediction models specific to construction
environments and production delay factors generated by those environments. Figure 3.4

depicts a prelim schematic structure of the framework that was developed.

 

 

 

 

 

 

 

 

 

 

SEM Collect Last Planner Data from
Literatui/ Construction Project(s)
I Identify Equation I I Identify Constraints J

 

 

 

 

A

Categorize Latent and ‘/—-I Record Measurements

Measured Variables

 

 

 

 

 

 

 

 

 

A

:I Specify Model I ConvertData to
usable format

 

 

Conversio
n

 

 

 

 

SEM
Softwar

 

 

 

 

 

L Identify Model ]

 

 

Matrices, Measurement

Errors
I Estimate Model J+——l
l

I Test Model I

l

I Modify Model I

 

 

 

 

 

 

 

 

I Report Results I

 

Figure 3.4 Sample representation of Final Framework

3.2.3.1 Framework Demonstration
Along the development of the framework, randomly simulated data will be used to

demonstrate its implementation. The data will be generated on the basis of the survey

-88-

instrument created (a deliverable) to collect the data in actual scenarios. The
demonstration will further help in uncovering potential roadblocks in the execution of the
process and will contribute to developing strategies and establish guidelines to
successfully implement the framework. For instance, the demonstration may reveal issues
such as model misspecification and identification issues or handling large sets of
variables. More importantly the results reveal whether the assumed relational equation is
appropriate for the observed data. SEM is a multi-approach design tool and
demonstration may reveal a requirement to explore different approach to design the
equation.

3.3 Chapter Summary

This chapter provided an overview of the approach that was adopted for achieving the
overarching research goal to improve workflow reliability at the production level in a
construction project. The brief outline of the research method illustrates how the different
fields of Structural Equation Modeling, Lean Construction techniques, and conventional
production management techniques were explored to develop a framework to establish a

relation between site performance (reliable workflow) and production delay factors.

-39-

CHAPTER-4

FRAMEWORK DEVELOPMENT

AND DEMONSTRATION

-90-

4.1 Introduction

After the Phase-1 of the research (refer Fig. 3.2), which is extensively discussed in
chapter-2, this chapter discusses the remainder of the 3 phases and presents the primary
deliverables of this research that include scope establishment, development and
demonstration of the framework for designing a relational equation. An extensive
literature analysis on production management systems and structural equation modeling
was conducted to develop the framework. The framework was primarily developed from
essential attributes of structural equation modeling and incorporating all essential
requirements for application in production management environments.

The framework builds on the current abilities of production management systems
and their metrics identified in chapter-2 and develop strategies to adopt or modify these
metrics to utilize them effectively for the relational equation development purposes. The
development of framework is accompanied with demonstration to facilitate the

understanding of the reader.

4.2 Scope of Framework Development

The extensive literature review on SEM revealed many attributes of SEM analysis. SEM
is applied to a variety of fields and as a result several techniques within SEM have
evolved to cater to analysis of data according to its nature and differing requirements.
SEM modeling is particularly useful to (I) investigate proposed cause-and-effect
relationships, and (2) conduct predictive analysis. The basic difference between these two
uses is the method of data collection. To perform prediction modeling, the measurements

of the MVs are taken repeatedly over time and their autoregressive effects are studied, bi-

-9]-

directional relationships between latent variables are accounted for as well. Cross-
sectional designs allow only for the evaluation of relationships among variables at one
point in time and do not allow for autoregressive effects or time lags, therefore, these
studies are conducted to study cause-and-effect relationships based on past experiences
and cannot dynamically predict the outcome. This research has focused on developing the
framework to produce cross-sectional equation designs because the first step towards
developing a prediction model is to establish a basic relational model between various
factors, the repeated measures design will become the next step in the sequence (not

developed in this research).

4.3 Framework Development

The following discussion will lay out the framework for developing a Structural Equation
Model specific to the problem of workflow reliability on construction sites. The
framework described will include demonstration at every step (if required). The model
developed as part of demonstration is based on information collected by the researcher by
doing literature reviews. The model thus created at the end of the framework will only
represent one of the many possibilities that arise in real world application, and should be
considered as an example model by the reader for future applications only at the their

own discretion/risk.

4.3.1 Phase-l State Hypothesis

The first step in building the model is establishing the hypothesis that is being tested or

confirmed in the model. “Theory must be the foundation of even the simplest of models,

-92-

because variables could always be linked to one another in multiple ways. Most would be
complete nonsense. Theory makes the model plausible.” (Hair et a1. 2005)

The hypothesis advanced by this research through an extensive literature review
is: Workflow reliability is impacted by production delay factors that come into effect
during construction. Thus, the researcher is interested in exploring one possible cause
and effect relationship between production delay factors and resultant site performance

(consistent site performance represents reliable work flow).

4.3.2 Phase-2 Identify the Constructs
The first step after underlining the theory is to identify the key constructs and establish
relationship between the constructs (Hair et al. 2005). The researcher is interested in
understanding the impact of production delay factors over work flow reliability. As
discussed in chapter 2 earlier, in the Last Planner system during the constraint analysis
stage, the needed inputs for an assignment are classified into three main categories
namely: Pre Requisite Work, Directives, and Resources. These are primarily phenomena
that if present, prevent the start of an assignment, that is, if there is a lack of pre-requisite
work, or directives, or resources, then the planned work is prevented from starting
(Ballard 1997). Therefore, these three categories assume prime importance in deciding
the outcome of the work and were selected to study their relation with reliable workflow.
Lean theory advocates the presence of three main phenomena that lead to delayed
or unsatisfactory completion of work. These are Muda (Unnecessary work/waste
generation), Mura (Variation), and Muri (Overburden). Site factors that lead to any of

these phenomena are instrumental in preventing a task from finishing on time and/or

-93-

desired quality. Finishing the task at hand is equally important as starting it and work in
process is not value added work until it is complete. Therefore, studying the impact of
factors that prevent work to finish is equally important as impact of factors that prevent
work from starting.

Thus, the key research question is: How do the key construction site phenomena -
Prerequisite Work, Directives, Resources, Waste, Variation and Burden - affect work
flow reliability? From literature review (Liker 2004; Ballard 1997) and researchers’
experiences, the researcher developed the following relationships:

0 ‘Availability of Prerequisite Work’ positively affects ‘Work Flow Reliability’.
0 ‘Availability of Directives’ positively affects ‘Work Flow Reliability’.

0 ‘Availability of Resources’ positively affects ‘Work Flow Reliability’.

0 ‘Waste Reduction’ positively affects ‘Work Flow Reliability’.

0 ‘Variation Reduction’ positively affects ‘Work Flow Reliability’.

0 ‘Burden Reduction’ positively affects ‘Work Flow Reliability’.

These six above stated relationships form the basis of how the researcher feels the
production factors influence work flow reliability on the construction site. After
relationships are specified, the next step is to identify the model in a form suitable for
analysis. This step involves identifying constructs as endogenous or exogenous, followed
by demonstrating the relationship visually in a path diagram (Hair et al. 2005), in other
words, the directional relationships are established between the constructs.

In the demonstration example, the listed relationships identify seven constructs.
Table 4.1 below indicates which constructs are endogenous or exogenous. The researcher

theorized that the factors that comprise each of these construction site phenomena are

-94-

independent of each other (although in field, it has been demonstrated that waste comes
from all other five constructs), therefore, each of these constructs representing all these
phenomena are exogenous constructs. However, work flow reliability is being studied as
an outcome of the presence of all these constructs, therefore, conceptually it is dependent
on the entire six constructs, and as a result it is depicted as an endogenous construct. In
addition, each construct is assigned an abbreviation for ease of repetitive mention of

constructs in the research and also for convenience of handling them in the EQS software

used for this research.

 

 

 

 

 

 

 

 

Name ol‘t'onstruct 'l‘ypc ol‘t'onstruct .A*\hbrc\i'.1linn
1. Work Flow Reliability Endogenous Construct WFR
2. Availability of Prerequisite work Exogenous Construct PRE
3. Availability of Directives Exogenous Construct DIR
4. Availability of Resources Exogenous Construct RES
5. Variation Reduction Exogenous Construct VAR
6. Burden Reduction Exogenous Construct BUR
7. Waste Reduction Exogenous Construct WTE

 

 

 

Table 4.1 Categorizing the constructs into endogenous and exogenous constructs (with abbreviation)

The constructs are identified as endogenous or exogenous constructs and
represented in a path model as shown in Figure 4.1. Because they are all latent
constructs, i.e., they cannot be measured directly and represent a larger condition created
by a myriad of factors, it becomes imperative to identify the variables that truly capture
the constructs by direct measurement. These variables are better known as indicators or

measured variables (Hair et al. 2005). For example, Variation in work cannot be

-95-

 

measured so accurately as to eliminate uncertainty, but by measuring factors that

influence or cause variation we can measure the variation to a greater accuracy.

@ D‘

Figure 4.1 Path Diagram of the proposed structural model (1” Stage)

Thus, the next step in the process is to define the constructs and Operationalize
them by selecting their measured variables and their scale type. Identifying scale type is
critical because it converts the qualitative data into quantitative data. The quantitative
data is required to develop the observed covariance matrix, however, it may be noted that
the scale is not standardized because of the reasons discussed in section 2.6.1 of chapter
2.

4.3.3 Phase-3 Define and Operationalize the Individual Constructs

This step begins with defining the constructs involved. The definition is necessary in
order to provide basis for selecting or designing individual indicator items. This step is
followed by operationalizing the constructs by selecting their measurement scale items
and scale types (Hair et al. 2005). This step often involves a series of scale items in a
common format such as Likert scale or a semantic differential scale, but in certain
circumstances it involves the use of composite scales. It can be done in two different

ways. The first approach is to define and operationalize constructs as they were in

-96-

previous research studies, using literature search on the individual constructs and identify
scales that previously performed well. The second approach is to develop new measures
because the theory under study does not have a rich history of previous research (Hair et
al. 2005).

The hypothesis in question does not have a rich history of previous research;
however the identified constructs are concepts that have been discussed in previous
research (Ballard & Howell 1998; Ballard 1997; Liker 2004) therefore, a mixed approach
was followed to define and operationalize the constructs. The constructs were developed
by populating them with measured variables and vetting with experts to rate how well the
definition and variables match the construct. A preliminary checklist of measured
variables was generated and binned under the identified constructs. The checklist was
prepared using a mix of previous research and researcher’s experience in the use of
constraints analysis on the field (the source for each item in checklist is provided in
Appendix C). The language of items sourced from previous search was modified to
present them in question format. The checklist was then sent to 5 experts in the field of
lean construction; the experts were asked to rate in a yes/no format against each item in
the checklist, whether the item suitably captures a production delay factor on the
construction site and contributes to the measurement of the construct it is classified into.
Based on the expert’s responses, the checklist items receiving majority of the votes (3 out
of 5) were retained and the rest were discarded. The variables and definition are listed
after receiving majority voting on them. The definitions are as follows:

a) Work Flow ReliabilitL (WFR): Work flow reliability concerns a state of

consistency, dependability, and predictability, and improving reliability generates a more

-97-

consistent, dependable, and predictable flow (Thomas et al. 2003). It is therefore, the
measure of consistency of flow of activities carried out for completion of a set of tasks in

order to achieve a larger production goal in adherence with a plan. Two Likert items are

provided as item indicators for this construct. These are:

   

Measured Variables Scale Scale Reference

WF R1. The PPC is high; 90% or higher 1-10 (IO-100% Agree)

   

  

 

 

 

WFR2. We are headed to a timely completion of the 1-10 (10—100% Agree)
project

 

 

 

 

Table 4.2 Measured Variables for Work flow Reliability Construct

Since both the indicators are reflections of the construct in question, that is, a high work
flow reliability can be represented by high PPC as well as high probability of timely
completion of the project instead of vice versa, these two indicators are reflective
indicators (refer chapter 2 for definition).

b) wilability of Prerequisite Work (PRE): As suggested by the name it includes
all the planning factors that make sure that the prerequisite work is available in time,
desired quality, and in time etc. It is difficult to measure precisely the ‘availability of
prerequisite work’ without measuring the production factors that together contribute to

availability of prerequisite work. The measured variables designed for this construct are:

-93-

\lezlsuretl \i'uriuliles Scale Scale Rclcrcncc

 

PREl. Prior work is complete with desired quality. 1-10 (10-100%
Complete)

PRE2. The pre-requisite work has been verified to meet 1-10 (IO-100% Agree)
current work needs (dimensions, locations, etc)

 

 

PRE3. All regulatory inspections needed were conducted 1-10 (IO-100% Agree)
and successfully passed.
PRE4. The prerequisite work is officially handed over 1-10 (IO-100% Agree)
from prior trade.
PRES. The weather forecast is favorable for performing 1-10 (IO-100% Agree)
work.
PRE6. Access to the work area (permits, etc) has been 1-10 (l0-100% Agree)
obtained and cleared with respective party.
PRE7. All coordination issues with following trade have 1-10 (IO-100% Agree)
been addressed by the project team.
PRE8. Overall, the prerequisite work is complete. l-10 (IO-100%
Complete)
Table 4.3 Measured Variables for Availability of Pre Requisite Work Construct

 

 

 

 

 

 

 

 

 

 

As discussed in chapter 2 earlier, each formative construct needs at least two
directional paths emitting from it to reflective variables or reflective constructs or a mix
of both. As depicted in Figure 4.1, each of the six exogenous constructs have a path
emitting from it to the WFR construct, which is a reflective construct. Therefore, all the
formative constructs require only one more path emanating from them to a reflective
indicator or a reflective construct. Mackenzie et al. (2005) recommended introducing a
global measure as a reflection of the construct itself to meet this model specification
condition. Hence, variable PRES is designed as a global measure to capture the overall
condition of pre requisite work, to fulfill the condition of model identification. All other

measured variables are formative indicators.

-99-

 

c) Availability of Directives (DIR): As suggested by the name, it includes all the
planning factors which take care of the procedural and technical requirements during
construction, for example, D&E documents, submittals, RFIs etc. It is difficult to

measure precisely the ‘availability of directives’ without measuring the factors that

contribute to it. The measured variables populated for this construct are:

\lCLlSUl'CLl Variables Scale Scale Reference
DIR]. The D&E documents for current work are l-10 (IO-100% Agree)
acceptable.

   

 

DIR2. The D&E documents for current work are 1-10 (IO-100% Agree)
available.

 

DIR3. All RFI's for current work are addressed 1-10 (IO-100% Agree)
satisfactorily.

 

DIR4. The owner is agreeable to any design/scope 1-10 (10-100% Agree)
changes, if any.

 

DIRS. The standard of work performance is available. 1-10 (IO-100% Agree)

 

DIR6. The standard of work performance is clearly 1-10 (IO-100% Agree)
expressed.

 

 

DIR7. The submittals are approved. 1-10 (IO-100% Agree)
DIR8. The submittals are available. 1-10 (IO-100%
Available)

 

DIR9. Final design instructions have been confirmed by 1-10 (IO-100% Agree)
architect/engineer to avoid change.

 

 

 

 

 

 

 

DIR10. All lien waivers are in order. 1-10 (IO-100% Agree)
DIRll. All permits are available. 1-10 (IO-100% Agree)
DIR12. All lien waivers are submitted. 1-10 (IO-100%
Complete)
DIR13. Overall, all the directives are in order. 1-10 (IO-100% Agree)

 

Table 4.4 Measured Variables for Availability of Directives Construct

The variable DIR13 is the only reflective indicator in the construct and is
designed solely for model identification purposes. All other measured variables are

formative indicators.

~100-

  

 

d) Availability of Resources (RES): This construct involves making sure that all
the resources; material, labor and equipment are available in right quantity, desired
quality and at the right time. It is difficult to measure precisely ‘availability of resources’
without considering possible factors involved in making the resources available. The

measured variables populated for this construct are:

\leaxurerl \"arialiles Scale Reference

 

RES]. The material is/will be delivered on time. 1-10 (IO-100% Agree)
RESZ. The material specifications match 1-10 (IO-100% Agree)
contract/submittal specs.

 

 

 

RES3. Space is available for material lay-downs. l-lO (IO-100%
Available)
RES4. Path available for material transport. 1-10 (IO-100%
Available)

 

RESS. The labor is informed of all work assignments. 1-10 (IO-100% Agree)

 

RES6. The labor needed is available. 1-10 (IO-100%
Available)
RES7. The tools to be used are in good working 1-10 (IO-100% Agree)
condition.
RESS. Right tools are available in enough quantity for 1-10 (IO-100%
crews to work with. Available)
RES9. The equipment to be used is in good working 1-10 (IO-100% Agree)
condition.
RESIO. Right equipment is available in the desired 1-10 (IO-100%

 

 

 

 

 

 

 

 

 

quantity. Available)
RES] 1. Overall, all the required resources are available. 1-10 (IO-100%
Available)

 

Table 4.5 Measured Variables for Availability of Resources Construct

The variable RESll is the only reflective indicator in the construct and is
designed solely for model identification purposes. All other measured variables are

formative indicators.

-lOl-

e) Waste Reduction (WTE): is defined as creating a production environment that
leads to least amount of waste (non-value adding work, poor quality work etc.)

generation. Since it is difficult to measure it precisely, several measured variables are

created to measure lack of waste more accurately. These measured variables are:

\leasuretl \I"arialiles Scale Scale Reference

WTE]. The amount of work is sized appropriately to l-10 (IO-100% Agree)
avoid overproduction.
WTE2. Transportation of raw/processed materials 1-10 (IO-100% Agree)
requires single handling (as opposed to more than once).

   

    

 

 

WTE3. Transportation of raw/processed materials is 1-10 (IO-100% Agree)
over short distances.
WTE4. No amount of rework/correction is required for 1-10 (IO-100% Agree)
completed work.
WTES. No over-processing (unnecessary finishing) is 1-10 (IO-100% Agree)
performed for completed work.
WTE6. Inventory is delivered just-in-time to the 1-10 (IO-100% Agree)
production unit.
WTE7. The production unit makes no unnecessary 1-10 (IO-100% Agree)
movements to complete the work.
WTE8. The production unit is always working to 1-10 (IO-100% Agree)
complete the work.
WTE9. The production crew is consulted on the best 1-10 (IO-100% Agree)
way to perform the work.
WTE10. The production crew is consulted on the safest 1-10 (IO-100% Agree)
way to perform the work.
WTEll. Overall, sufficient steps have been taken to 1-10 (IO-100% Agree)
prevent generation of waste.

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.6 Measured Variables for Waste Reduction Construct

The variable WTEll is the only reflective indicator in the construct and is
designed solely for model identification purposes. All other measured variables are

formative indicators.

-l02-

0 Variation Reduction (VAR): is defined as creating a production environment that
decreases the possibilities of variation in production capacities and performance of the

crews. Since it is difficult to measure it precisely, several measured variables are created

to measure variation reduction as accurately as possible. These measured variables are:

\leasured Variables Scale Scale Reference

VARl. There was no absenteeism in the workforce. 1-10 (IO-100% Agree)

   

    

 

VAR2. The production crews operate with reserve 1-10 (IO-100% Agree)
capacity.
VAR3. The learning curve effect is realizable for the 1-10 (IO-100% Agree)
crews.
VAR4. Recovery plans from occupational accidents 1-10 (IO-100% Agree)
are standard practice.
VARS. The production rates of the crew are consistent 1-10 (IO-100% Agree)
over time (match Takt).
VAR6. The production plan has workable backlog in 1-10 (IO-100% Agree)
case of interruptions to current assignment(s).
VAR7. The production plan has time buffers to prevent 1-10 (IO-100% Agree)
interruptions to project completion.

 

 

 

 

 

 

 

 

 

 

Table 4.7 Measured Variables for Variation Reduction Construct

All except VARS are formative indicators, since they contribute to form the
environment for reduction in variation. The VARS variable is a reflective indicator
because it is a result of Variation Reduction that the production rates of the crew are
consistent over time.

g) Burden Reduction (BUR): is defined as maintaining a production environment
that reduces the physical, cognitive and psychophysical burden on the crew and
equipment to avoid burnout and stoppages as a result. Since it is difficult to measure it
precisely, several measured variables are created to measure burden (or lack of it). These

measured variables are:

-lO3-

Measuretl Variables

Scale

Scale Reference

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BURl. The work environment is free of potential 1-10 (IO-100% Agree)
environmental hazards (air-borne pathogens, dust,

chemical agents, ultra-violet light, and ionizing

radiation).

BUR2. The work environment is comfortable to function 1-10 (IO-100% Agree)
in (temperature and humidity, lighting, sun exposure,

noise levels).

BUR3. The work environment is not congested. l-10 (IO-100% Agree)
BUR4. Overtime is not needed on our jobs. 1-10 (IO-100% Agree)
BURS. The work is adequately paced to avoid physical 1-10 (IO-100% Agree)
fatigue.

BUR6. The workers are equipped with PPEs, as l-lO (IO-100% Agree)
necessary.

BUR7. The workers are not allowed to engage work 1-10 (IO-100% Agree)
involving heavy muscular loads.

BUR8. The workers have access to material-lifting 1-10 (IO-100% Agree)
equipment.

BUR9. A Job Safety Analysis is performed. 1-10 (10-100% Agree)
BUR10. Workers are trained on proper lifting l-lO (IO-100% Agree)
techniques.

BURll. Rest cycles are built into the work method. 1-10 (lo-100% Agree)
BUR12. Workers receive sufficient information 1-10 (IO-100% Agree)
regarding the process flow and output.

BUR13. The rate of information does not exceed the l-10 (IO-100% Agree)
mental capacity of the worker to process.

BUR14. Identical or very similar signals don’t occur for 1-10 (IO-100% Agree)
a long time.

BUR15. Adequate time is allowed for decisions and 1-10 (IO-100% Agree)
resulting actions in the normal circumstances.

BUR16. Adequate time is allowed for decisions and 1-10 (IO-100% Agree)
resulting actions in emergencies.

BUR17. Hand tools used are the correct ones for the 1-10 (IO-100% Agree)
task.

BUR18.Hand tools used are adequately maintained. 1-10 (IO-100% Agree)
BUR19. All safety signs and visuals are correctly located. 1-10 (IO-100% Agree)
BUR20. Overall, the workers do not feel any kind of 1-10 (IO-100% Agree)

burden.

 

 

 

Table 4.8 Measured Variables for Burden Reduction Construct

-104-

 

The variable BUR20 is the only reflective indicator in the construct and is
designed solely for model identification purposes. All other measured variables are
formative indicators.

After operationalizing the constructs, the next step is to validate the constructs.
Construct validation is performed to verify “the extent to which a set of measured items
actually reflects the theoretical latent construct those items are designed to measure”

(Hair et al. 2005).

4.3.4 Phase-4 Construct Validation
There are four components of the procedure to establish construct validity. Three of those
components namely convergent validity, discriminant validity and nomological validity
(partially) are tested once the model is estimated and factor loadings (parameter values on
directional paths) are calculated.

Convergent validity is established by measuring all the factor loadings on the
paths of the construct. The minimum statistically significant loading is a value above .7
because square of that value is around .5, explaining half the variation in the item with
the other half being error variance. This validity tests stands good for reflective
constructs, because in theory the indicators of the reflective constructs closely represent
the latent variable and are interchangeable. Therefore, if the factor loadings of those
variables are less than .7, then they are not converging and therefore not valid. However,
these similar thresholds are not applicable to formative indicators, since formative
indicators can influence even 0.1 of variance in the item and yet be an important part of

the construct. Therefore, convergent validity is not checked for our example in question.

-lOS-

Discriminant validity checks the uniqueness of each construct as compared to
another in the complete model. This is tested by comparing the variance-extracted
percentages of any two constructs with the square of the correlation estimate between
these two construct. Since the latent construct should explain its item measures better
than it explains another construct, the variance extracted percentages should be greater
than the squared correlation estimate. Hence, applying the same logic of formative
indicators as in convergent validity test, this test is not an effective test for formative
indicators.

The other two validation tests are nomological validity and face validity.
Nomological validity assess the degree to which the construct as measured by a set of
indicators predicts other constructs that past theoretical and empirical work says it should
predict. In the current example, the theory in concern has implemented a lean thinking
and proposed an entirely new set of constructs than the ones used in past research.
Therefore, the constructs proposed in this research lack nomological validity and stand as
suspects with the burden of proof lying on the researcher proposing the new constructs.
They can be validated only after estimating the structural model.

The final test of validity is called face validity, and is the most important test of
validity because it tests theoretical soundness of the variables in a construct (Hair et al.
2005). This validity is done based on researcher’s judgment and expert opinions. It has
already been mentioned earlier in the chapter in section 4.3.2 that the constructs were
populated with their respective variables based on vetting by expert judges after an
extensive literature review. Therefore, it is reasonable to say that the constructs pass the

test of face validity.

-l06-

After validating the constructs, SEM analysis can move forward in two different
ways. First approach is to specify and estimate the measurement model followed by
specifying and estimating the structural model. The second approach involves directly
specifying and estimating the structural model. The following section presents a

discussion on the first approach.

4.3.4.1 Develop the Overall Measurement Model

The main purpose of developing a measurement model is to specify the relationship
between measured variables and their constructs. This step involves creating a path
model between the constructs and their measured variables and also estimating the
parameter loadings on those paths. However, the relationships between different
constructs are kept bi-directional (to indicate covariance) specifically to strictly test the
measurement theory of the model. By testing measurement theory, means, statistically
identifying the relationships between constructs and the measured variables and testing
for measurement model fit. This involves specifying the path model in the sofiware
program and estimating the model using the collected data. Figure 4.2 shows the

measurement model for the example used in this research.

-l07-

cum.

 

.2 E..£%_.'=«$ e~§
_. m

.‘rw‘ $2

I.

,-
a:

3‘

m

 

 

 

 

.~\
,,,,,,,,,/r/xf

Vll

      

    
      

[I

// / .I.//ll

/ , .x...../ wwm Okll'

/,..,...,./m_o of lV
xwmo 071V

-.
-illxll mum oblllV
lll lea BlalV

- flux-d oplv

lllllwmm Okl

mun. o...

 

mum Elf
ma opr

l~=o Okllllll

Figure 4.2 Diagram showing the constructs in a measurement model relationship

-108-

_ :mmm ‘ FREE :wwm

     
    

A ll-

 

l l . l . .u ........ \.

logo--3? . .
Al ll lmu>>o._..l. ..

1A.... H... w- - 5..- e-.- .

All 1 :mEs Bl ll l. l. .. l. . l
A- l llm<> op: . ll 1 . .l l l l. .l.

All l .lll o lllll l. ..... l l l l -l llull:
l as. c..- l - - ll l-.. l. -5- .-

PNw

 

3:0

NED

Figure 4.2 Cont‘d,

-l09-

After the model is specified and estimated, the factor loadings between construct
and variables and the error variance terms are known entities. Therefore, their values are
fixed to the loading estimates obtained from the measurement model before the structural
model is estimated. Due to change in the relationships between constructs (from
correlational to directional) during transformation from measurement to structural model,
the loading estimates are subjected to change, thus, changing the fit of the model. This
change signifies difficulty in sorting out which indicator item measures a particular latent
construct (Hair et al. 2005). This situation is called interpretational confounding and
generally results from instability associated with under-identified factors.

In addition, in a measurement model, given that constructs are in a bi-directional
relationship, the identification of each construct becomes harder especially in a formative
construct. Recalling the guidelines of identification of a formative construct from chapter
2, a formative construct needs at least two directional paths into reflective indicators or
constructs or a mix of both. Meeting this requirement is difficult when there are no two
(at minimum) theoretically sound reflective indicators for a construct. Therefore,
especially for formative models this approach to SEM analysis is discouraged. This
research recommends and adopts the second approach to conduct structural model
analysis.

Since the framework does not support measurement model testing, the next step is

to develop an overall structural model.

-llO-

4.3.5 Phase-5 Develop and Specify the Structural Model

Specification of a structural model is an extension to specification of the measurement
model. In the measurement model, only relationships between constructs and their MVs
are specified. and the constructs are specified as correlated. However, in the structural
model, the relationships between various constructs are specified as originally proposed
in the theory. As mentioned earlier, six relationships were hypothesized as part of the
theory, these were:

H1: ‘Availability of Prerequisite Work’ positively affects ‘Work Flow Reliability’.

H2 = ‘Availability of Directives’ positively affects ‘Work Flow Reliability’.

H3 = ‘Availability of Resources’ positively affects ‘Work Flow Reliability’.

H4 = ‘Waste Reduction’ positively affects ‘Work Flow Reliability’.

H5 = ‘Variation Reduction’ positively affects ‘Work Flow Reliability’.

H6 = ‘Burden Reduction’ positively affects ‘Work Flow Reliability’.

These hypotheses mean that structural relationships between constructs would be
specified according to the dependence relationships mentioned, i.e., a directional arrow is
emanating from all of the six constructs towards work flow reliability. In addition, the
directional relationships between the constructs and their measured variables are also
specified (if measurement model is not already developed).

The structural model designed per the hypotheses in the example is shown in
Figure 4.3. The model diagram is designed by using the Diagrammer function of the EQS

software.

-lll-

Note: It is advised to create the diagram of a model only after the measured variables are
defined in the Data file of EQS. Otherwise, the data file and the diagram file will not
coordinate during the Estimation procedure of the model.

The first step after developing the structural model is to specify the structural
model. Specification involves ‘setting the scale’ for the latent variable. Because it is
unobserved, a latent factor has no metric scale, meaning no range of values. This scale
can be provided in two ways:

1. A scale can be set by fixing one of the factor loadings and setting its value. It is
recommended that the value is set to 1.
2. The construct variance can be set to a value. Again, 1 is the recommended value.

However, method 2 is not applicable to formative constructs, since the software
reads formative constructs as dependent variables; it does not assign them a variance of
their own. Further, in case of formative constructs, it is recommended that the factor
loading on the path from the construct to the reflective variable (added for identification
purposes) since it is conceptually a representation of the construct itself.

In Figure 4.3, the structural model for the example, the red arrows indicate the
fixed factor loading on the reflective variable of each construct. For the WFR construct,
since both its MVs are reflective, one can choose any one variable and fix its factor

loading to 1. Once the model is specified it is checked for identification.

-l12-

 

REES ’ " RESZ RES1 Er

 

 

 

 

 

 

 

E72
W
@
REES \
Rese fries—1d] RE‘S11 W517 er1o WTE9
El.
FRE?‘ VAR‘l
PRES
PRE4 @ @v VAR4. E48
m ,
‘PRE'I
-fi—ee
a
g
g “ any»:
9‘ \\.\ Bums
.. a
E70

Figure 4.3 Diagram showing the Structural Model of the proposed Hypotheses
- 113 —

4.3.5.1 Model Identification
The first step in model identification is to check whether enough information is provided
by the sample covariance matrix to estimate all the parameters. This is calculated by
counting the number of data points and the number of parameters to be estimated in the
model. The equation for data point calculation is
Number of data points = p(p+1)/2, where (l)
p is the number of measured variables. In the example, p = 72. Therefore, data points =
2628. The number of parameters to be estimated can be counted from figure 4.4 (count
the number of asterisks).
The number of asterisks present in the figure are:
72 loading estimates (ignore the fixed loadings for this purpose)
+ 8 error variance terms (ignore the fixed loadings for this purpose)

+ 6 exogenous-endogenous structural terms

 

= 86 free parameters

The difference between the data points and parameters to be estimated gives the
df of the model, which in this example is 2628 - 86 = 2542. Since the df is more than
zero, the model is considered over-identified (which is the ideal condition).

The second condition to check for model identification is rank condition and
requires the researcher to verify algebraically whether two equations can define any
dependent variable. If such a condition exists then it means it violates the rank condition
of the model. Such an error can be associated with linear dependence error message, post
analysis. In formative constructs or dependent latent variables, setting the disturbance

term is a delicate issue.

-ll4-

The disturbance term represents all remaining causes of the latent variable other
than the set of indicators which form the respective latent variable. It is suggested by
Diamantopoulos (2006) that setting the disturbance term to zero is legitimate practice as
long as the researcher is theoretically sure that all possible causes of the construct are
included as indicators in the model. Not setting the disturbance to zero also has reported
to create the rank condition error, but this practice should not be used simply to overcome
rank identification issues.

The third condition for model identification is the MIMIC modeling rule for
formative indicators. As discussed earlier, each formative construct needs at least two
paths emanating from it to reflective indicators or reflective constructs. This condition is
met for the example, as indicated in figure 4.4. The next step after developing a structural
model is to design a study and collect data for model estimation purposes which is

discussed in the next section.

-l15-

E48

.- m .7. a .
mun“: J ,. .
m u;.;. “EWWWMW m

     

           

    

BBB

 
     

rum

-ll6-

0N
THEE «3.3 ii a
x

\\

Figure 4.4 Structural Model showing fixed and free parameters.

    

1

 

 

PRE8

/.
. . . ow...
, , . . 5:6

4.3.6 Phase-6 Design Study to Produce Empirical Results

Designing a study involves several aspects like deciding on sample size, selecting the
estimation method, and selecting a method to handle missing data. Since the data to be
collected needs to be quantitative in nature, each measured variable is accompanied with
a measurement scale like ordinal scale, interval scale, categorical scale etc. The selection
of an ordinal scale is most common practice for example, Likert 0-10 scale or semantic
differential scale etc. A mixture of scales within the variables of a single construct is
welcomed if necessary to capture the variable in its truest form, since, mixing scales only
increases the software processing and results interpretation time but it causes no problem
in estimating the model. Similarly, having different scale ranges for variables only come
with a risk of extra computational time.

The scales used for the example in research are Likert scales with a range of 0-10
value. The scales against each variable are shown in Tables 4.3 to 4.8. None of the
measured variables have a physical quantity to measure; therefore, no real scale of
measurement has been assigned. This is primarily because all of the variables signify a
planning measure each and can only confirm or report completion of that measure on

some relative scale.

4.3.6.1 Determining Sample Size

SEM is a study involving large sample sizes. “The role of sample size is to produce more
information and greater stability, which assists the researcher in performing SEM. Once
the researcher has exceeded the absolute minimum size i.e. one more observation than the

number of observed covariances, larger samples mean less variability and increased

-n7-

stability in the solutions. As teviewed in the literature, several considerations are taken
care of while determining the sample size for a model. These are:

l. Multivariate Distribution: As data deviate from the assumption of multivariate
normality, the impact of sampling error tends to increase on the results of the model. To
counter these non-normality effects, the ratio of respondents to parameters need to be
higher. Expert opinions regarding these ratios, to minimize the problem with deviations,
are varied. Hair et al. (2005) cited a ratio of 15 respondents/ parameter; whereas, Kline
(2005) suggested that an ideal case would be 20 cases/ parameter but anything less than 5
cases/ parameter would probably not yield trustworthy results. Raykov et al. (2006), also
suggest 10 cases/ parameter as a guideline.

2. Estimation Technique: The most common procedure to estimate SEM models is
Maximum Likelihood Estimation (MLE) method. MLE is an iterative approach that
makes small sample sizes more likely to produce invalid results. Kline (2005) have
suggested an absolute minimum sample size to have any hope of getting good results
would be N = 100, while 100-200 would be a medium sample size, and over 200 would
be generally considered a large sample. Hair et al. suggest a sample size in range of 150-
400, but subject to some more considerations like number of constructs in the model. It is
recommended that if the factors (latent variables) are larger than six then the sample size
can exceed 500. However, MLE method is sensitive to extremely large amount of data
and almost any difference is detected, making the GOF measures suggest poor fit (Hair et
al. 2005). The estimation method used in the example is also MLE.

MacCallum et al (1996) provided a method for determining the minimum sample

size necessary to achieve a given level of power for tests of model fit, a discussion on

-118-

those steps is beyond the scope of this research. Therefore, the researcher suggests
reading MacCallum et al. (1993 and 1996) for a detailed discussion on topic of power
analysis and sample sizes.

Factors like missing data also need to be considered in determining the sample
size. If more than 10% missing data is expected than higher sample size is needed to
overcome the missing data (for a detailed discussion on the 10% missing data as
threshold, the researcher suggests reading Hair et al. 2005).

As highlighted in the discussion it is evident that the researcher should be
informed about the number of parameters to be estimated to calculate a sample size. The
sample size determined after taking all considerations into account for the example in
demonstration is 400. The next step after determining sample size is to prepare for
analysis of missing data, as it is rare to find a set of data especially involving large

sample sizes to not have missing data problem (Hair et al. 2005).

4.3.6.2 Handling Missing Data

“Missing data can have significant impacts on any analysis, particularly those of a
multivariate nature” (Hair et al. 2005), such as SEM. Missing data can occur due to
reasons like data entry errors, data collection problems or even refusal to answer by the
respondent to a particular or a set of questions. Missing data has both practical and
substantive impact on data analysis; the practical impact being that in multivariate
research, missing data may eliminate so many observations that an adequate sample may

be reduced to inadequate sample. From a substantive perspective, in case of non-random

-ll9-

missing data (a pattern of missing data can be seen), statistical results can be biased and
lead to erroneous results (Hair et al. 2005).

Missing data can be handled using several approaches. The approach to handle a
missing data depends on the (1) nature of the missing data, i.e., if the data is Missing
Completely at Random (MCAR), or Missing at Random (MAR), (2) extent of missing
data, i.e., what amount of data is low enough for the analysis to produce erroneous
estimate, (3) the type of missing data, i.e., whether it is ignorable or not ignorable data
(Hair et al 2005). After all these determinations are made, the method of imputation for
missing data is selected, if required. Considerable amount of literature is present on
handling missing data; again, a topic outside the scope of this research. Further, the
example used in this research consists of randomly generated hypothetical data set,
therefore, no missing data exists to present as an example. The reader is referred to one
excellent source for handling missing data (Hair et al. 2005).

The next step after designing the study to produce empirical results is to actually

collect the data. The following section presents a discussion on data collection.

4.3.7 Phase-7 Data Collection

Collecting data requires identifying the proper sample population. It is important to keep
the theory in mind while selecting the sample population; bad population will lead to bad
responses, which may lead to bad fit of the model. For example, the theory proposed in
the research is concerned with studying the causal relation between site production
factors and reliable work flow. The appropriate population group for collecting data is

construction industry professionals who function as a member/leader of the management

-120-

team and use/ have experience with the Last Planner system. Given the interest in
collecting responses based on enough experience with the Last Planner system and
reliable work flow, the requirement was set for at least 1 year work experience or 1
project experience in that role.

However, due to the limitations of finding such a large number of professionals
meeting the experience requirement, the example demonstration had to be estimated
using randomly generated hypothetical data. The data was generated using the
RANDBETWEEN command in Microsoft Excel for 400 points for each of the 72
variables, with possible answers on the scale 1-10.

Since the data is hypothetical, no real conclusions can be drawn from the SEM
analysis on the example, rather conceptual conclusions will be drawn in the spirit of

demonstration of the modeling process and its benefits.

4.3.7.1 Multi-Collinearity
The next step after collecting the data is to check for multi-collinean'ty in the constructs.
In formative constructs, high multi-collinearity can destabilize the model. The Variance
Inflation Factor (VIF) statistic is used to determine if formative indicators are too highly
correlated. Scholars, suggest that VIF values greater than 3.3 indicate high multi-
collinearity (Roberts and Thatcher 2009).

The VIF factors for constructs of the demonstration model were calculated using
MINITAB software using the regression analysis command and selecting the VIF output.

The results for construct “Availability of Prerequisites” are presented in Table 4.9.

-lZl-

 

Regression Analysis: PRE1 versus PRE2. PRE3,...

The regression
equation is

PRE1 = 5.21 - 0.0407 PREZ - 0.0108 PRE3 — 0.0465 PRE4 + 0.144 PRES -
0.0144 PRE6 + 0.0016 PRE7 + 0.0382 PRE8

Predictor
Constant

PRE2

PRE3
PRE4
PRES

PRE6

PRE7
PRE8

Coef

5.2147
-0.04073

-0.01078
-0.04651
0.14388

-0.01438

0.00165
0.03816

5 = 2.94846

SE Coef
0.7699

0.05372

0.05406
0.05195
0.0507

0.05338

0.05246
0.05147

R-Sq =

T
6.77
0.76
-0.2
-0.9
2.84
0.27
0.03
0.74

2.5%

p
0

0.449

0.842
0.371
0.005

0.788

0.975
0.459

VIF

1.036

1.05
1.008
1.04

1.039

1.027
1.03

R-Sq(adj) =

0.7%

 

Table 4.9 Multi-collinearity test for Prerequisite Construct using MINITAB

As indicated in Table 4.9, the multi-collinearity results for indicators of

the validity of the proposed formative constructs.

4.3.8 Phase-8 Assess Structural model validity

-l22-

demonstrated in the next section, using the example.

“Availability of Prerequisite” construct are between 1.008 to 1.05 which are well below
the threshold of 3.3. Similarly, the VIF results for all other formative constructs were

below 3.3. Thus, in the hypothetical scenario, multi-collinearity does not pose threat to

The next step is to estimate the structural model and test its validity. This step is

The structural model shown in figure 4.4 can now be estimated. The procedure to run the

model in EQS is briefly discussed to facilitate the understanding of the software (for

detailed discussion on using EQS, refer Byme 2006, an excellent source on this topic).
After the path model is created using the diagrammer and model is specified by fixing
appropriate loadings like one factor loading per construct and all error and disturbance
loadings. The data is entered in the data type file of EQS using raw data format and not in
form of covariance or correlation matrix as it cannot produce robust statistics using the

matrices (Roberts and Thatcher 2009).

 

,
EQS Model Specifications

 

 

EQS Model Specifications I - I
EQS Model Title ‘ .
muffin} Jr- Em- " LE ._ . -_ !

InputData Filelniormation ~- - -- W __.._.-,-_.-

I file Info I c:\eqs61\thesis 400 samples. ass

I Variables - I72 Cases - I400

 

 

 

 

 

l Type olAnalysis _- - ' :Advanced Options

I [— Multisample Analysis Categorical variables I

I I 9.1; . - [IT—— I .

. r— Structural Mean Analysis Missing data handling I

I . ‘ I l

._ F-W'tlévemréws- Misc. Options | .
l Normal theory estimators D I I I
l (- LS (- GL5 (- ML_ l eete cases I

~ Non-normal estimators corrections [Select one item only] —~~ - ~-

I I'- Elliptical l— AGLS
; l— Heterogeneous kurtosis [— Robust methods

(.7 .- F 6* i-_ _E F . s - ;‘

Cancel I

 

 

 

 

 

 

Figure 4.5 EQS window showing selection of Estimation method

When on the diagrammer file, care should be exercised that the data file is open in
the background, click Build EQS> Title/Specifications to create the command file for the
software. The ‘model specifications’ window as depicted in Figure 4.5 appears and
requests for type of estimation method. The best practice recommends, as mentioned

before, selecting MLE method along with robust methods to overcome non-normal

-l23-

estimators. After clicking ‘Ok’, the command file is generated which looks like Table
4.10 below (note that this command file is created for the example used in this research).

The figure only shows a partial file, for complete file refer to Appendix B in the end.

 

 

/TITLE

Model built by EQS 6 for Windows

/SPECIFICATIONS

DATA='C:\eq561\thesis 400 samples.ess';

VARIABLES=72; CASES=400;

METHOD=ML,ROBUST; ANALYSIS=COVARIANCE; MATRIX=RAW;

/LABELS

V1=PRE1; V2=PRE2; V3=PRE3; V4=PRE4; V5=PRE5; V6=PRE6; V7=PRE7; V8=PRE8;
V9=DIR1; V10=DIR2; V11=DIR3; V12=DIR4; V13=DIR5; V14=DIR6; V15=DIR7;
V16=DIR8; V17=DIR9; V18=DIR10; V19=DIR11; V20=DIR12; V21=DIR13;

V22=RESl;; ......

/EQUATIONS
V8 = lFl + E8; V21 = 1F2 + E21; V32 = 1F3 + E32; V43 = 1F4 +
E43;
V48 = 1F5 + E48; V72 = *F7 + E72; F1 = *Vl + *V2 + *V3 + *V4 +
*VS + *V6 + *V7; F2 = *V9 + *V10 + *Vll + *V12 + *V13 + *V14 + *V15
+ *V16 + *V17 + *V18 + *Vl9 + *V20 ;F3 = *V22 + *V23 + *V24 + *V25 +
*V26 + *V2? + *V28

/VARIANCES

Vl = *; V2 = *; V3 = *; V4 = *;

/COVARIANCES

/PRINT

EIS;

FIT=ALL;

COVARIANCE=YES;
TABLE=EQUATION;
/OUTPUT
Information matrix;
Parameters;

Sigma;

Covariance matrix;
Standard Errors;
Listing;
DATA='EQSOUT.ETS';
/END

 

Table 4.10 Command file generated after selecting estimation method

On the command file, go to Build EQS > Output to select various outputs required

fi'om the estimate. The Output options window is as shown below in Figure 4.6.

-124-

 

The researcher can choose to store all or selected results from the window. After
the output options are selected the model is estimated by clicking Build EQS > Run EQS.
The estimation output is generated in an output file format. Refer to Appendix B for
complete results from the model run. The overall fit statistics for the model are reported

in the Table 4.11.

 

r
, Build Output Options fil

 

l— Store derivatives
l l— Store gradient Cancel I
l [— Store inverted information matrix

I l— Store parameter estimates

l— Store model covariance matrix [SIGMA]

l— Store sample covariance matrix

I l— Store standard errors

I l- Store weight matrix

I l— Store standardized solutions

‘ l— Store results from LMtest

l— Store results from Wtest

l— Store Bentler-R aykov corrected Fl-square
l— Store univariate statistics of the input data
I I— Write a codebook for information stored in the paragraph

f7 Store all of the above results
I: List EQS results in the log file

17 Data file name [eosourers

 

 

 

 

 

 

Figure 4.6 Window showing Output options for Model Estimation Results

-125-

 

 

MULTIVARIATE KURTOSIS

 

MARDIA'S COEFFICIENT (G2,P) = -86.9022
NORMALIZED ESTIMATE = -8.4185

GOODNESS OF FIT SUMMARY FOR METHOD = ML

INDEPENDENCE MODEL CHI-SQUARE = 3399.033 ON 2556DEGREES OF FREEDOM

INDEPENDENCE AIC = ~1712.967 INDEPENDENCE CAIC = -14471.151
MODEL AIC = -l939.475 MODEL CAIC = -l4338.273

CHI-SQUARE = 3028.525 BASED ON 2484 DEGREES OF FREEDOM

PROBABILITY VALUE FOR THE CHI-SQUARE STATISTIC IS 0.00000

THE NORMAL THEORY RLS CHI-SQUARE FOR THIS ML SOLUTION IS
2755.168.

FIT INDICES

BENTLER-BONETT NORMED FIT INDEX = 0.109
BENTLER-BONETT NON-NORMED FIT INDEX = 0.335
COMPARATIVE FIT INDEX (CFI) = 0.354
BOLLEN'S (IFI) FIT INDEX = 0.405
MCDONALD'S (MFI) FIT INDEX = 0.506
JORESKOG-SORBOM'S GFI FIT INDEX = 0.839
JORESKOG-SORBOM'S AGFI FIT INDEX = 0.830
ROOT MEAN-SQUARE RESIDUAL (RMR) = 0.423
STANDARDIZED RMR = 0.051
ROOT MEAN-SQUARE ERROR OF APPROXIMATION (RMSEA) = 0.023
90% CONFIDENCE INTERVAL OF RMSEA ( 0.020, 0.026)

RELIABILITY COEFFICIENTS

CRONBACH'S ALPHA 0.058

 

Table 4.1 l Goodness-of-F it Statistics for the demonstration model

4.3.9 Phase-9 Interpreting the Results

As recommended in the literature, an SEM should be tested for fitness using three
indices; one basic goodness-of-fit index, one badness-of-fit index, and one incremental fit
index. The most commonly used fit indices are Chi-Square Index, RMSEA index, and
CFI Index, in that same order. Table 4.12 displays threshold values for the GOF

rneasures.

-126-

 

$080203.
.38 00 50 .00: ”00.30mvmcomfigm 300... 2.0.0.500 $0.00 0.590-305.0000 w505m00E00 000%:— gm 500.0030 .50 0059.50.00.20 N. .v 030,—.

 

 

 

 

«020 0.0500. «00.002030 «0 000.00"). ..m0.00...0> \0 .0025: "E “0.02
.0030 .0 .0030 .0 .0090 .0 3. .003...
om. .0 Eu 53> mm. .0 Eu 5.2. mm. .0 Eu 53. 0600 Eu 5.2. .0 mm.u_.u 5.2. .0090 .0 ENEU
3v m03m> 3v m0:_m> 3V m0:_m> mo.v m0:_m> 8v m020> 5:... wo.v m0:_m> 4.09).”.
30%...
.050 0m: 3m. C0090 00:05 .050
Rm. 0600 EU fim. 0600 EU “0.02.0: 00220 0600 Eu 5;: .0 mm. .0 EU 0m: 0.02.03
533 $2 .0 mo. 533 $2 .0 mo. 00 2:00 mo. E...5 3.0. 5.2.. 3.2 .0 mo. 003.0 00 Baou 22%
:02. mm
08..” A z 804 A z 80; A z 5:... 00500500035

53> 0m: 50: 00 5:... 0m: 50.. 00 0.0.... 50: 00 50 0350.0

.2. .8. $6.... .2. .Nm. 9.2.... 03.0.. .0 mm. Na. 26.2 .052. .0 mm. .9. >02 2..

om. 0>00< ~m. 0>o0< .0500 .0 mm. Nm. 0>00< .0500 .0 mm. .0500 .0 mm. Sh .0 Eu

.2... 000m 5 000m .53.

00509.0 00509.0 53. 5.3.0. 005080 :0>0 :30.
00 :00 m0:_m> 00 :00 33:3 :8 m02m> 00 :8 333 :3 m0:_m> 00509.0 m0:_m>
.0 280205. -a 28:20... a 2.80.5.2. a 2.80.53 .0 280.55 .0 $80.50... 0.83-20

0%... omvsvfi Name 0%... omvevu a... .5. 2.0.3...

m0_0m_.m>
om~ A2 8N v z .0 .oz

 

—127-

l. The Chi-Square Index for the model is 3028.525 with df= 2484, p = 0.000. As
discussed earlier, since the p-value is lesser than .05, the chi-square value is statistically
significant which means the model does not have a good fit. Interpreted literally, this test
statistic indicates that given the present data, the hypothesis bearing on Work flow
reliability causations, as presented in the model, represents an unlikely event and should
be rejected (Byme 2006). Since, the sensitivity of chi-square index to sample size has
already been identified in the literature review; it is necessary to run a few more fitness
tests to reach a conclusion.

2. The comparative fit index (CFI) of the model is reported as 0.354 which is
much less than the recommended 0.9 or higher (refer table 4.12). Therefore, the model
does not pass the CF I test.

3. The RMSEA value for the model is reported as 0.023 with a 90% confidence
interval of RMSEA (0.020, 0.026). The RMSEA value is lower than the recommended
0.07 value. Therefore, the model passes the RMSEA (badness-of-fit) test.

The model under testing failed two out of three fit tests, therefore, it can be
concluded that the model hypothesis can be rejected for the chosen sample population.
However, it is important to note that the GOF measures such as Chi-square index are
sensitive to sample size (as explained in chapter 2). For example, in this scenario, if the
sample size was 4000, the chances of achieving a good chi-square fit would still be
remote because the Maximum Likelihood Estimation (MLE) procedure would be thrown
off. Traditionally, there is a trade-off sample size with the GOF. In this example, the

trade-off sample size was not determined.

-128-

 

 

Covariance Matrix to be Analyzed (Trimmed)

BUR16 BUR17 BUR18 BUR19
BUR20
V66 V67 V68 V69
V70
BUR16 V66 7.788
BUR17 V67 0.424 8.295
BUR18 V68 0.507 0.294 8.195
BUR19 V69 -0.377 0.194 0.155 8.017
BUR20 V70 -0.341 -0.145 -0.215 0.142
8.090
WFRl V71 0.050 0.042 0.739 0.379
0.258
WFR2 V72 0.228 —O.267 -0.797 0.175
0.346

BENTLER-WEEKS STRUCTURAL REPRESENTATION: (SPECIFICATION STATUS)
NUMBER OF DEPENDENT VARIABLES = 15
DEPENDENT V'S : 8 21 32 43 48 70 71 72
DEPENDENT F'S : l 2 3 4 5 6 7

NUMBER OF INDEPENDENT VARIABLES = 73

INDEPENDENT V'S : 1 2 3 4 5 6 7 9
11

INDEPENDENT V'S : 12 13 14 15 16 17 18 19
22

INDEPENDENT V'S : 23 24 25 26 27 28 29 30
33

INDEPENDENT V'S : 34 35 36 37 38 39 40 41
44

INDEPENDENT V'S : 45 46 47 49 50 51 52 53
55

INDEPENDENT V'S : 56 57 58 59 60 61 62 63
65

INDEPENDENT V'S : 66 67 68 69

INDEPENDENT E'S : 8 21 32 43 48 70 71 72

INDEPENDENT D'S : 7

NUMBER OF FREE PARAMETERS = 144
NUMBER OF FIXED NONZERO PARAMETERS = 16

3RD STAGE OF COMPUTATION REQUIRED 957288 WORDS OF MEMORY.
PROGRAM ALLOCATED 40000000 WORDS

DETERMINANT OF INPUT MATRIX IS 0.20969D+63

PARAMETER CONDITION CODE (IDENTIFICATION STATUS)
D7,D7 LINEARLY DEPENDENT ON OTHER PARAMETERS

10

20

31

42

54

64

 

Table 4.13 Covariance matrix (trimmed) and model specification & identification results.

-l29-

 

Other than the fit indices results, the output file also generates the covariance
matrix and summary of model specification and identification issues, as shown in Table
4.13. The specification status confirms the researcher’s specification of dependent and
independent variables, as well as the free and fixed parameters. In identification status,
the ideal message is “Parameter estimates appear in order, No special problems were
encountered during optimization”. It is important to locate the message prior to any
interpretation of results (even for GOF indices). One of the most common errors, as
shown in Table 4.13, is the condition code “Linearly dependent on other parameters”.
This situation occurs because either the parameter is underidentified in the model or it is
empirically underidentified as a consequence of the data. Strategies on handling these
errors are discussed in a later section of this chapter. Note that as a result of the error
mentioned, the results of this model are not correct, however, they are being discussed to
help the reader to interpret the results.

The primary focus of the estimation process in SEM is to yield parameter values
such that discrepancy (i.e. residual) between the sample covariance matrix S and the
population covariance matrix 2, i.e. (Z — S) is minimal. Values greater than 2.58 are
considered large; the lower the value of residuals, the better the hypothesized model fits
the sample data. It should be noted that EQS produces two results; residual covariance
matrix, and standardized residual matrix. It is recommended to use the latter result to
avoid misinterpretation due to the unit of measurement of the observed variables (Byme

2006). An example of standardized residual matrix is shown in Table 4.14.

-l30-

 

Standardized Residual Matrix

BUR16 BUR17 BUR18 BUR19 BUR20
V66 V67 V68 V69 V70
BUR16 V66 0.000
BUR17 V67 0.053 0.000
BUR18 V68 0.063 0.036 0.000
BUR19 V69 -0.048 0.024 0.019 0.000
BUR20 V70 0.001 -0.006 0.000 -0.010 -0.004
WFRI V71 0.009 0.006 0.091 0.045 0.029
WFR2 V72 0.029 —0.032 —0.097 0.022 0.042
WFRl WFR2
V71 V72
WFRl V71 0.087
WFR2 V72 -0.042 0.001
AVERAGE ABSOLUTE STANDARDIZED RESIDUAL = 0.0389
AVERAGE OFF-DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL = 0.0399

LARGEST STANDARDIZED RESIDUALS:

NO. PARAMETER ESTIMATE NO. PARAMETER ESTIMATE
1 V72, V8 0.572 11 V30, V4 0.147
2 V72, V14 0.462 12 V29, V10 ~0.147
3 V35, V15 0.174 13 V34, V10 -0.l46
4 V71, V13 0.170 14 V28, V13 —0.144
5 V64, V42 0.157 15 V38, V24 0.143
6 V69, V15 0.157 16 V45, V7 .-0.l42
7 V47, V18 -0.150 17 V36, V27 -0.141
8 V65, V18 -0.l49 18 V53, V6 0.141
9 V65, V29 -0.148 19 V62, V30 0.140

10 V15, V1 0.147 20 V49, V28 0.140

 

 

Table 4.14 Standardized Residual matrix (Trimmed) and largest standardized residuals
The data in Table 4.14 indicates that the largest standardized residual is 0.572,

and the average standardized residual is 0.0399, both values reflect good fit to the data. A
further review of frequency distribution (refer Appendix B) reveals that 97.79% of
residual values fall between -.1 to .1. This confirms that the model as a whole appears to
be quite fitting.

Assessment of residual covariances and model fit can only reveal information
about how close the model as a whole is to the data. However, assessing individual

parameter estimates is equally critical so as to make sure that the parameter estimates

-l3l-

 

exhibit correct sign and size and are consistent with the underlying theory. Any estimates
falling outside the range indicate that either the model is wrong or the input matrix lacks
sufficient information. A parameter estimate is said to be unreasonable if it shows the
following characteristics (Byme 2006):

a. Correlations > 1.00

b. Negative Variances

c. Covariance or correlation matrices that are not positive definite.

Also, the model is considered a poor fit if the standard errors are excessively large
or small. However, since standard errors are a function of units of measurement in
observed and/or latent variables, therefore, it is difficult to establish ‘1argeness’ or
‘smallness’ of the value (Byme 2006).

A sample of the data generated on parameter estimates by EQS is shown in Table

4.15.

-l32-

 

MEASUREMENT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH 0.

PRE8 =V8 = 1.000 F1 + 1.000 E8
DIR13 =V21 = 1.000 F2 + 1.000 E21
WFR2 =V72 = .030*F7 + 1.000 E72

.074
.403

CONSTRUCT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

F1 =F1 = -.001*V1 + .079*V2 - .001*V3 - .005*V4
.004 .052 .005 .005
-.262 1.523 -.219 -.863
- .003*V5 + .003*V6 + .001*V7
.005 .005 .004
-.591 .635 .207

VARIANCES OF INDEPENDENT VARIABLES
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

 

V F
V1 - PRE1 8.756*I I
.620 I I
14.124@I I
I I
V2 - PRE2 7.819*I I
.554 I I
14.124@I I
I I
V3 - PRE3 7.829*I I
.554 I I
14.124@I I
I I
V4 - PRE4 8.135*I I
.576 I I
14.124@I I
STANDARDIZED SOLUTION R-SQUARED
PRE8 =V8 = .076 F1 + .997 E8 .006
DIR13 =V21 = .004 F2 + 1.000 E21 .000
RESll =V32 = .163 F3 + .987 E32 .027
WTEll =V43 = .182 F4 + .983 E43 .033
F1 =F1 = -.015*V1 + .997*V2 - .012*V3 - .059*V4
- .036*V5 + .039*V6 + .012*V7
F2 =F2 = -.129*V9 - .008*V10 - .076*V11 + .044*V12
+ .972*V13 - .065*V14 + .007*V15 + .071*V16
- .135*V17 + .016*V18 + .050*V19 - .030*V20

 

Table 4.15 Standard Errors, Variances of Independent Variables, and Standardized Solution (Edited).

-133-

 

The data in Table 4.15 is presented in order of unstandardized estimates followed
by standardized solutions both for measurement equations and the variances (not reported
completely in the Table 4.15, for complete reference see Appendix B). The
unstandardized estimates are presented in the order of:

WFR2 = V72 = .030*F 7 (estimated value) + 1.000 E72
.074 (Standard error)
.403 (Test Statistic (Visible only if significant))

The standardized solutions are presented in a single line along with a related R2
value, as shown below:

PRE8 = V8 = 0.076 F1 + .997 E8 .006 (R2)

The R2 value indicates the proportion of variance of the response variable
accounted for by its predictor variables. It may be noted that in standardized solution,
parameters that were previously fixed to 1.0 take on new values; also, variances of
independent variables cease to exist because they take on the value of 1.0 (Byme 2006).

In reviewing the standardized estimates, the researcher should verify that
particular parameter values are consistent with the literature. The test statistic in the
unstandardized solution represents the parameter estimate divided by its standard error
and operates as a z-statistic in testing that the estimate is statistically different from zero.
Based on a level of .05, the test statistic needs to be > i 1.96 (or greater than two times
the standard deviation value, if checking for a different a level) before the hypothesis
(that the estimate = 0) can be rejected. Non-significant parameters, with the exception of

error variances, can be considered unimportant to the model in interest of parsimony

(simplification of model) (Byrne 2006).

-134-

The above discussion brought out the salient aspects of SEM modeling, EQS
software results and their interpretation. The model estimation results revealed that good
fit (in this case bad fit) alone is insufficient to support a proposed structural theory. The
researcher also must examine the individual parameter estimates against corresponding
predictions or paths, each representing a specific hypothesis. The methodology indicated
that during the development and assessment of the model, the most critical aspect was to
achieve a proper identification of the model. As the literature review pointed that the
model identification issues are recurring in nature and manifest both in the theoretical
aspect and computational aspect of modeling, this research proposes strategies based on
previous research and expert opinions to remove those problems in context of formative
construct modeling.

4.4 Discussion of Strategies to Assist in Model Development

Some of the common concerns expressed in previous research regarding the formative
modeling approach are model misspecification of formative constructs (Mackenzie et a1.
2005; MacCallum et al. 1993).

Firstly, the disturbance term, as previously discussed, is that part of the construct
that is not explained by its measured variables but which impacts the variable
(Diamantopoulos 2006). Therefore, to add a disturbance term to a formative indicator is
considered a standard part of specification. However, EQS requires constraining of error
and disturbance term variances to zero or 1 during the estimation process. This
requirement ofien leads to error messages such as: (1) Constrained at Upper Bound, or
(2) Constrained at Lower Bound. One approach to remedy this problem is to remove the

disturbance term from the construct altogether. This practice is not a strongly advised

-l35-

practice because it essentially mean that the researcher has high confidence that the
chosen formative indicators for the construct completely explain the variance of the
construct, which is rarely the case. Thus, this practice needs a sound theoretical
justification in order to be implemented to fix the problem.

Secondly, since formative constructs are essentially transformed into multiple
indicators multiple causes (MIMIC) models for identification purposes; one of the
methods to identify those models is to have one path to a reflective indicator and one path
to a reflective construct (a well suited method when the model configuration supports it).
However, having two paths in that configuration are not enough in many circumstances,
therefore, it is recommended to have at least two reflective indicators to the formative
construct. This is advantageous because: (a) the formative construct is identified on its
own and can go anywhere in the model; and (b) one can include it in a Confirmatory
Factor Analysis (CFA) model and evaluate its discriminant validity and measurement
properties. However, the disadvantage in this method is that it forces the researcher to
design an extra variable when initially it did not exist (Jarvis et al. 2003). In addition, it
uses up extra degree of freedom of the model and may affect the sample size in case there
are many such constructs.

The strategies above were discussed to support the model identification and
specification process of formative models. For further discussion on improving the model
identification for formative constructs, the reader is encouraged to refer sources like
Jarvis et a1. (2003), MacCallum et a1. (1993), and Mackenzie et a1. (2005).

The strategies above help in overcoming the recurring and fi'ustrating problems in

estimating SEM models, especially, the formative type models. Overall, the framework

-136-

for this research was deveIOped based on the literature review and demonstrated to
facilitate understanding. The following section presents a graphical representation of the

developed framework in Figure 4.7.

-137-

 

 

 

 

 

 

 

 

 

 

 

f EISVHJ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ecmcgsom .ootmécmémgu .«0 288332 com xHOBoEE... coco—goo 5v oSwE
v ngm C...
E n 395
€86.50 8 a
, - 529 68:52 as 32588 as .: V.85 .m as; 8
A 89:20:28
mam $9865 a. “2588 you ESE 5.2— @2035 .N
_ duo—Emcee
- i 28 $3865 5233 E56328 ”338$. . _
mHUDEmZOU mm... mNE<ZOF<MmEO
mZOEEO _
.55me _
. 828m 8088332 goo—om .m
v.88—
/_ . _ 28505830 Edam ._
0388ch 8 com: on 9 was: 05 @2060 . HZWEDEmZH
\F mHUDMHmZOU A<DQ~>~Q7= ME mZEm—Q >m>m3m 29me
l damage—2 Baboficmomsao 05 E
mail— ?Ewwfii: on 8 25: 3:38:00 06 >352

mHUDEmZOU mE EHZm—n:

 

 

 

 

6053982: on 2 Q3832»: 05 @283
mummEOm>m mH<Hm

 

hlzllgfi. m
ammz

 

4.5 Developed Framework

 

-138-

 

 

I I C'ISVHJ I Z HSVHJ l

8.288 E. as»:
e mm<=m On.

a

- . 303.588 “:82 .8 23m 05 @8on .m

63¢ .886:—

mdm can 153— 8900 conga—Eon. .28:— 50 x020 .m
— 8.3on0 5233 8595322 5.33%.. _

4

mQOE A<MDHODMHM >50QO QZ< m04m>m0

 

 

 

 

 

 

 

 

 

 

  
   

2.: .5
32:25.0

o..<

m mm<2m
OH ¥0<m 00

 

 

  
 

 

   

 

 

 

_ £23
3am can .momwoBEoZ 83:50

.ucmcgeoma 4590250 82:80
_ t53<> HODMHmZOU v.0mm0

E53<> H03MPmZO0 ¥0m10

 

 

 

£2; 8"; Ba @8382 _

 

 

 

  

 

 

ml<ll:Z_E
\lummmm
wane—.3360
be 0.532

 
 

 

 

   

0380:3— 03368...—

m mm<=h 20¢...—

 

S ElSVI-Id

7 EISVHJ

 

 

 

-l39-

:{88 S. 23m:

w mm<=m OP

:5qu HO: much—0 .:

 

.252
05 SEE

A

 

 

 

 

832 5:33:52
Lo; 55:. 6553.34

 

 

 

59:0 E maotm .8 0.020 .v
333% E330 .m
mla w .209: 2: :3. .N
3:930 85% Loo—om .—
me02 ”it. mhfizimm

 

 

 

 

 

 

 

 

 

H535 Soto:

 

 

 

as 8228.: fl

l 3030— .: u>_ 00 P. ..O .
952:2 llll b . .1 c 33935 N
\mmam

 

 

\/

 

 

 

 

 

>tm<mZEAOUAHJDE mOL MUMIU

I

EELS 3.: E Sec .25» .m
wldd , 52% E20 $8533 5230 ._ \

 

 

 

 

 

 

 

cozmocfii.
66:28.: :0 women
2886:. u>oEum 6 3E

 

 

 

\7

EBSEE 63368 .8 m.mv h:> ._ l

.256: 6:660:

 

<._.<Q PDQ—400

a.

 

 

 

 

 

 

 

 

 

sowoaq< Sac wEmfl—z .m
Amt—Ev 60522 :ocmEzmm goo—ow .N
ofim oEEmm cwfioo .—
whqumm

      

a
2mm

 

A<0§E2m m0DDO~E 0% >095 < 205mm

 

 

=55:me .333

 

a

m mw<=m SOME

:wioo ESL

 

L ElSVHcI

9 EISVHJ

 

 

|

.140-

.......l
...l

“ECW—

262500 S. 8:?

 

.momwouabm 258882 use
33mm: 2: Ea...“ Sofia—6:8 26.5

mHADmmM NE Hmdmdmhzh

 

 

3 gm 8 xommv 8% Bu:
5.3 “web was 382 05 250m

 

 

 

”UHOEE—wn— “Guava—ammo: baa —gaogm
.0 sum 23 Sega 68855 Ba .80 as 38?.

FED—:55 1500—2 mmmmm<

 

A

b Hm<=m SOME

 

6 SISVHJ

8 EISVHJ

 

-141-

4.6 Chapter Summary
This chapter discussed the SEM development framework in the context of the research
objectives. The survey questionnaire prepared to measure the production delay factors on
a construction site was also discussed. Further, a demonstration development example for
studying impacts of production delay factors on work flow reliability was presented. The
chapter also provides insight into some strategies for handling specification and
identification issues with SEM.

The example used for framework demonstration was based on the hypothesis that
“All production delay factors impact workflow reliability”, albeit with a hypothetical
data. The purpose of the detailed demonstration was to provide guidance for interpreting
the results and their statistical significance once a model is developed. The demonstration
illustrated the use of the EQS software and MINITAB in order to develop the SEM and
analyze the results. In the demonstration, it was found that departure from standard
methods of developing an SEM, as found in literature, is acceptable due to the nature of
the proposed model studies, for purpose of exploring the relationship between production
delay factors and reliable workflow.

Overall, this chapter presented the main deliverables, framework and survey

questionnaire, as formulated from the objectives of the research.

-142-

CHAPTER-5

SUMMARY AND CONCLUSIONS

-143-

5.1 Introduction

Chapter 5 provides an overview of the research conducted and its major findings along
with a brief discussion of how the goals and objectives were accomplished. This chapter

also discusses the limitations of this research and offers suggestions for future research.

5.2 Research Overview

This research provided a framework for studying relationship between production delay
factors and work flow reliability which was constructed based on literature review and
input fiom experts of Lean Construction and demonstrated by an example. To reiterate,
the last four chapters are briefly discussed below explaining how the framework was
constructed.

Chapter 1 presented the current level of implementation of production
management and control tools and understanding of work flow reliability for the purposes
of (1) improving production planning effectiveness and as a result (2) improving work
flow reliability. Based on the need statement, the chapter stated the overall goal and
objectives of this research along with its scope, and the potential benefits that could be
achieved on accomplishment of the research goal.

Chapter 2 provided an overview of current production management practices
discussing the conventional and lean production management tools. Existing researches
on production factors affecting construction projects were discussed as part of the
literature analysis in partial fulfillment of research goal and objectives. From the
literature of productivity delay factors, chapter 2 provided an insight into types and

classification of productivity factors. The Structural Equation Modeling analysis

-144-

technique was extensively reviewed and its nuances were highlighted. The need and prior
applications of SEM in construction management were discussed. Overall, the chapter
brought forth the essential attributes which were used to develop the intended framework
to explore the relationship between work flow reliability and production delay factors.

Chapter 3 discussed a four phased methodology that was adopted to accomplish
the research goal and objectives based on the literature review. The contribution of each
phase in achieving the research objectives was discussed. The chapter underscored the
need for reviewing literature in categories of conventional and lean production
management tools, productivity delay factors in construction and SEM analysis
technique.

Chapter 4 presented the main contribution of this research as it developed the
survey instrument to collect data for analysis of the impact of production delay factors on
work flow reliability. The literature review in chapter 2 and the first part of chapter 4
contributed in achieving the first research objective, which was to document literature on
production management tools, review the identified production factors in previous
research and develop a list of production delay factors to understand work flow
reliability. In addition, the chapter explained a step by step development of a framework
for constructing an SEM to capture the relationship between work flow reliability and
production delay factors. The framework was demonstrated using the actual survey
instrument to generate hypothetical data. Finally, the chapter discussed strategies to

overcome problems in developing the framework.

-145-

5.3 Research Goals and Objectives

The overall goal of this research was to understand cause and effect relations related to
workflow reliability at the production level in a construction project. In order to achieve
this goal, the research focused on developing a method to investigate the causal
relationship between production constraints and their impact on workflow reliability as
measure by the Percent Plan Complete (PPC) metric. To accomplish this two objectives
were proposed in chapter 1, which are:

1. Study production management tools implemented at the site level and
document the production delay factors encountered on construction site.

2. Develop a method to study the relationship between production delay factors
and work flow reliability.

Objective 1 was achieved by reviewing the literature of conventional and lean
production management tools and production delay factors in construction. The literature
review provided the essential attributes that were considered for developing a framework
to study the production delay factors affecting work flow reliability. These attributes
were: production planning metrics (such as PPC), production delay factors that occur on a
typical construction site, and classification of the production delay factors based on their
impact on start or finish of a construction activity.

Based on’the attributes, the research created a survey instrument to collect data on
production factors affecting construction site performance. The data would be used to
populate models that would be developed using the same production factors in the survey

instrument.

-l46-

Objective 2 was achieved by conducting the literature review on SEM, identifying
the essential attributes of SEM model design. Combined with literature on production
management, a fi'amework was constructed that allows studying the impact of production
delay factors on construction workflow reliability. The primary ingredients in developing
this framework were the production delay factors, which were mostly identified by
literature review and expert inputs, and specifics of SEM design, as adapted to
construction production management. The developed framework suggested number of
phases to be performed to (1) develop factors relevant to a construction company’s
specific sector and methods and (2) to establish a relationship model with these factors
and reliable .work flow by collecting data empirically and using SEM.

Each phase of the framework was demonstrated it with an example based on a
hypothesized model with a randomly generated hypothetical data. Software applications
like EQS and Minitab were integral to the development of the framework; therefore,
sofiware input/output was an integral component of the framework demonstration. The
framework could be applied to build SEM models for testing other relationships in the
field of construction management or build a different model based on envisioned

relationships (refer section 4.1).

5.4 Conclusions and Inferences

This section discusses the conclusions drawn from the literature review of production
management tools and observations made during the development of the framework. The
literature of conventional production management tool brought forward the different

metrics used to measure construction site performance. Discussion on tools like CPM and

-l47-

EVM highlighted their limited ability to control the project progress because they do not
provide any indication about the status of the upcoming work, which further limit their
ability to identify appropriate corrective actions before problems are encountered. The
SV and CV metrics implemented in the EVM method were criticized for not being
explicitly connected to appropriate management action (Walt 2006). Therefore, the
construction managers have to rely on their intuition to take corrective action. It is clear
from the implementation of SV and CV that these metrics only consider cost and
schedule as performance indicators, meanwhile, no reference is made to production
factors that actually impact the site performance. Therefore, these metrics fail to address

the root of the problem.

The literature review on lean production management tools revealed the use of
PPC as a metric to measure the reliability of production planning as implemented under
the Last Planner system. As discussed in chapter 2, PPC is a better indicator of site
performance than CPM and EVM, as it gives a better indication of competency of
planning the work. The constraints analysis in the Last Planner system provides a
window to look at the factors that are affecting the construction site performance.
However, it does not work with a standardized and exhaustive list of production factors
that affect the construction performance, instead, it records the factors such as, (1) late or
defective materials, (2) prerequisite work incomplete, (3) changes in priorities, (4)
absenteeism or accident (manpower), (5) failure to accurately estimate the amount of
labor time required to execute assignments, etc. These factors come from occurrences as
reported by the person conducting constraint analysis. Therefore, it is subject to

misinterpretation regarding root causes affecting the construction performance, which

-l48-

further affects selection of a corrective action. As a result, it is necessary to populate a list
of factors, which cover most, if not all, aspects of production planning and execution; in a
checklist format. Also, no attempt has been made at the production level of construction
to study the individual and collective influence of all the production delay factors that

affect construction performance vis-a—vis work flow reliability.

The reviewed literature on productivity factors in construction revealed an array
of efforts from the researchers to document factors that affect labor productivity; some
studied the impact of individual factors on construction productivity (Oglesby et al.
1989). However, this research revealed the importance of studying the collective impact
of those factors, because the model estimation revealed different values when three big
factors were assessed for their impact on reliable workflow versus the complete set of
factors. Consideration of multiple factors will help the production planners to strategize
more effectively in a real world complex and dynamic construction site.

In addition, the review revealed that there can be no standard list of production
delay factors that can be treated as a finite list, instead, the production factors can change
according to the practices of the construction company, type of construction project, size
of the construction project, etc. The list of factors should be selected by production
planners utilizing industry experiences. The survey instrument designed in this study
presents list of production delay factors relevant to most construction settings.

The SEM literature review, applications of SEM, using a framework revealed that
having a robust measurement theory is as important as the structural theory itself, i.e.,
proper care should be taken in deciding directional relationship between latent and

measured variables. This research has concluded that the latent factors (Pre Requisite

-l49-

Work, Directives, Burden, Variation, Waste, and Resources) are mostly subject to
constructivist interpretation, i.e., they form as a result of collection of a set of measured
variables and represent a collective existence of those variables, therefore, it is
recommended that future research consider and test relationships with formative latent
variables. This is especially important in a research of this nature because several
production factors that are not inter-correlated come together to represent a broad
planning category (latent variable), which, if tested for validation for reflective
relationships would have to be removed from the construct and affect the nature of the
latent variable. Although, from the measurement point of view, it is essential to add a
couple of reflective variables to each formative construct to validate the construct. In this
regard, the developed framework presents a methodology specific for conducting SEM

analysis as it relates to reliable workflow.

5.5 Limitations of the Research

The demonstration of the SEM framework was restricted to randomly generated data.
The reason behind this limitation is attributed to the large sample requirements required
for SEM analyses (min. sample size 200). Given that the survey instrument required
collection of data from construction management professionals based on their past
experience and familiarity with Last Planner system, it was impossible to find that
number of professionals. Therefore, the proposed structural theory (see section 4.3.1 and
4.3.2) between production factors and work flow reliability remains untested. Also, as a

result of using a uniform random variable, 400 realizations of the variable, any likely

-150-

correlation between the measured variables was not exhibited that could otherwise exist
with a real data.

In addition, the research did not discuss handling missing data prior to estimating
a measurement model, because the issue of missing data presents a vast discussion in
itself and there were no missing data in this case.

Further, the framework development focused on cross-sectional design in SEM.
Cross-sectional designs allow only for the evaluation of relationships among variables at
one point in time and do not allow for autoregressive effects or time lags, therefore, cause
problems in inferring causality or directional influence in cross-sectional studies. To
consider such an inference as valid, it would have to be assumed that the time lag during
which causal influence operates is essentially instantaneous, thereby justifying concurrent
measurement of variables in a cross-sectional design (MacCallum and Austin 2000).
Nevertheless, this research has focused on developing the framework to produce cross-
sectional equation designs, due to extreme complexities, stronger statistical background,
and longer time commitment required to pursue a framework for repeated measures
design. Moreover, this research was exploratory, which further justifies the use of cross-

sectional design.

5.6 Research Benefits and Contribution

As identified during the literature review of this research, reliable work flow as measured
by PPC has been typically low for many types of construction projects (Ballard 1999).
This is attributed to lack of understanding amongst the practitioners about the quantity of

impact each production factor has on reliable workflow. This research has developed a

-151-

framework which industry professionals can apply to overcome this overarching problem
and design models for their establishments and measure the impact of those factors that
affect work flow reliability. Those models can be actively implemented by project
planners/ Last Planners on the project to study the impact and improve work flow
reliability on the projects.

Although there have been several prior research in construction literature
attempting to study the impact of various factors on several issues like labor productivity,
performance on international projects, success of joint ventures, and project performance,
etc., no study to assess the impact of production delay factors on work flow reliability
was found. This research contributes to the Lean Construction practices by developing a
list of production delay factors that can be added to the constraint analysis sheet of the
Last Planner and used to record the status of production on a quantitative scale, on a
regular basis. The use of scale will assist in a more accurate cognition of the presence of
factors delaying the work.

In addition, the research delivered a framework that will not only assist in
studying work flow reliability, but it will also contribute in developing relationship
models for management practices at any organizational level in construction
organizations. For example, studying the impact of estimation/ bidding practices over
project award success.

Further, in chapter-l, the researcher pointed out the need to study the impact of all
factors instead of only a selected few on any phenomenon was pointed out based on prior
research (Maloney 1981; Oglesby et al. 1989). To test the validity of the statement,

another SEM was created to study the impact on work flow reliability using only 3 latent

-152-

factors (formative constructs), namely, propensity to reduce waste, burden, and variation.
The model was pretested for validity similar to that used for demonstration. The EQS
output of the results of the model after estimating with same set of hypothetical data are

presented in Table 5.1.

 

 

GOODNESS OF FIT SUMMARY FOR METHOD = ML

INDEPENDENCE MODEL CHI-SQUARE = 797.425 ON 780 DEGREES
OF FREEDOM
INDEPENDENCE AIC = -762.575 INDEPENDENCE CAIC = -4655.9l7
MODEL AIC = -734.466 MODEL CAIC = -4433.l42
CHI-SQUARE = 747.534 BASED ON 741 DEGREES OF FREEDOM
PROBABILITY VALUE FOR THE CHI-SQUARE STATISTIC IS .42602

THE NORMAL THEORY RLS CHI-SQUARE FOR THIS ML SOLUTION IS
717.505.

FIT INDICES

BENTLER-BONETT NORMED FIT INDEX = .063

BENTLER-BONETT NON-NORMED FIT INDEX = .605

COMPARATIVE FIT INDEX (CFI) = .625

BOLLEN'S (IFI) FIT INDEX = .884

MCDONALD'S (MFI) FIT INDEX = .992
JORESKOG-SORBOM'S GFI FIT INDEX = .918
JORESKOG-SORBOM'S AGFI FIT INDEX = .909

ROOT MEAN-SQUARE RESIDUAL (RMR) = .385

STANDARDIZED RMR = .047

ROOT MEAN-SQUARE ERROR OF APPROXIMATION (RMSEA) = .005
90% CONFIDENCE INTERVAL OF RMSEA ( .000, .016)
RELIABILITY COEFFICIENTS

CRONBACH'S ALPHA = -.015

 

Table 5.1 Goodness-of-fit Results for waste-variation-burden model

The goodness-of-fit results clearly indicate that the model with only waste-

variation-burden factors have different fit values for the same data. The RMSEA and CFI

-153-

 

values are significantly different, indicating that the model is less likely to be bad
(Hypothetically) at the same statistical significance as the full model. Further, a

comparison can be drawn by looking at the parameter estimates of a few common

variables in both models. The comparison is shown in Table 5.2.

 

 

 

Variable Demonstration Model Waste-yariation-hurden
result Model result
BUR20 .205 F6 + .979 E70 .075 F6 + .997 E70
WFR] .701 F7 + .713 E71 .229 F7 + .973 E71

 

 

 

Table 5.2 Comparison of parameter estimates (Standardized solution) of same variables in two models

Here, F6 represents the latent construct ‘Burden’ and E70 represents the disturbance error
associated with it. And, F7 represents the latent construct ‘Workflow Reliability’ and E71
represents the disturbance error associated with it.

The results clearly indicate that factor loadings and error variances for both
variables differ in the two models, which will lead to different interpretations by
production planners regarding the strategy to adopt to control those two production
factors. The difference comes about because the collective impact of the universe of
indicators differs in both cases. Thus, the experiment confirms the need for studying the
impact of all factors together as compared to isolation studies.

In addition, as discussed in previous chapters, it is advisable to study management
practices in a formative approach; this fiamework provides an opportunity to researchers
to revisit earlier studies, discussed in chapter 2, and study the impact of factors on the
respective end result in a manner more suitable for construction management practices.

This research will primarily benefit construction contractors especially those who

implement Last Planner system of production control, with a survey instrument for

-154-

 

standardized constraint analysis. Also, the list of production factors is designed in
accordance with the principles of lean production control. This framework will aid
contractors to bring structure in their evaluation of construction site performance through

the lens of production delay factors applicable to their practices.

5.7 Future Areas of Research

The introduction of SEM analysis technique in construction management research is still
in a nascent stage and merits a huge potential for implementation in studying cause-and-
effect relationships in this field, as was discovered during the literature analysis and
framework development and demonstration stage.

Much of the applied SEM literature is characterized by inadequate understanding
or acknowledgement of the limitations of single studies. Most often conclusions are
limited to the particular sample, variables, and time frame represented by the study. The
results are subject to sampling or selection effects with respect to at least three aspects of
a study: individuals, measures, and occasions. The choice of individuals has an effect on
sampling results, in order to account for such effects researchers may use expected cross-
validation index (ECVI), which is computed from a single sample, as an index of how
well a solution obtained in one sample is likely to fit an independent sample (MacCallum
and Austin 2000). Therefore as a first potential research area, actual data should be
collected and the proposed model in this research should be tested for fit. Also, a good fit
does not imply a universally true model for a set of factors; therefore, further research is
required to produce models with better fit to the data and hence a better prediction

capability. In addition, the list of factors may not be entirely applicable to any project and

-155-

only represent the pool suited to researcher’s purpose, hence, more research is required to
be done to refine the list of factors to study work flow reliability.

Also, during the development of the model, effort was made to include the entire
checklist of important factors that influenced production delay; as mentioned earlier, this
checklist was created by reviewing previous research, vetting with Lean experts and
researcher’s experience. However, a large model indicated the limitations of the
computational capabilities of a standard user computer assembly (normally found in
construction company offices) and a potential data handling and data processing problem
during the data collection phase, therefore, further research can be done to achieve model
parsimony (both from a statistical and logistical perspective). Model parsimony can be
achieved by field testing the model and performing statistical procedures like Factor
Analysis or Analytical Hierarchy Process (AHP).

Secondly, as discussed earlier in SEM literature review, there are two types of
equation design. The designed framework assists researchers to study the impact of
factors in a cross sectional design format, i.e., studying the factors and their impact at a
single moment in time. However, construction production is a dynamic phenomenon;
work flow reliability is tested on a weekly basis and is cumulative in nature, therefore,
reiterating the point expressed in the limitations section of the research, establishing
predictive capability merits further SEM research using the repeated measures design
aimed at studying the cumulative impact of factors on work flow reliability and vice
versa over the entire duration of the project. Such a study will be especially helpful to
create models designed to understand production planning effectiveness for each type of

construction project classified according to the building characteristic such as industrial,

-lS6-

commercial, healthcare, nuclear, etc. Accordingly, further research is required to modify
the framework to assist in developing models with repeated measures of all variables.
Construction is a global activity, and is executed by people from different cultures
all across the globe. Since, culture provides a behavioral context especially when there is
a huge presence of human interaction, it is important to examine the weighted value of
production factors that involve direct human involvement, such as burden, work flow
variation in this case. Therefore, a third potential area of research is to investigate the
extent to which such factors carry more weightage in overall performance and assign

those weights and study the relationship in cultural context as well.

5.8 Chapter Summary

This chapter discussed how the research goal and objectives were achieved, summarized
the accomplishments in previous chapters, and presented the final conclusions of the
thesis. The chapter also suggests three major areas for further investigation in production

management of construction projects.

-157-

APPENDICES

-158-

APPENDIX A

Survey Instrument - Participant Consent Form

DEVELOPMENT OF METHODOLOGY TO UNDERSTAND CAUSE AND EFFECT
RELATIONSHIP BETWEEN PRODUCTION DELAY FACTORS AND
CONSTRUCTION SITE PERFORMANCE

Principal Investigator: Tariq Abdelhamid
Secondary Investigator: Samarth Jain

The Michigan State University Center for Construction Project Performance Assessment and
Improvement is conducting a research project to explore the relationship between production
delay factors and construction site performance. The research will study the current efforts in
construction industry to maintain reliable workflow on construction work site and identify the
production factors that have an effect on construction performance. The research will develop a
method to study the cause and effect relationship that will assist construction professionals to
build their company specific models to increase their understanding of their own company
construction operations. Funding is being provided by the Construction Project Performance
Assessment and Improvement, School of Planning Design and Construction, Michigan State
University.

As part of the research we are interviewing professionals who are involved with construction
projects as a part of the management team. As an experienced industry participant, your insight
into the onsite construction production factors will be very useful to attaining the aims of this
research.

As a participant in this research, you will be asked a series of closed ended questions relating to
construction onsite management through an online survey. Your participation is voluntary and
you may choose to terminate your involvement in this study at any time during this project. If
you are uncomfortable at any time during the questioning, you may terminate and withdraw from
the interview. You may refuse to answer any particular interview question. Your privacy will be
protected to the maximum extent allowable by law. However, your title (e.g. Project Manager)
will be reported. The estimated time to complete this survey is approximately 30-40 minutes.

If you have any questions about this project, you may contact Dr. Tariq Abdelhamid, School of
Planning, Design and Construction, Michigan State University at (517) 432-6188, or Samarth
Jain, School of Planning, Design and Construction, Michigan State University at (517) 203-9010.
If you have any questions or concerns about your role and rights as a research participant or
would like to obtain information or offer input, or would like to register a complaint about this
research study, you may contact, anonymously if you wish, Michigan State University Human
Research Protection Program at 517-355-2180, FAX 517-432-4503, or e-mail irb@msu.edu. or
regular mail at: 202 Olds Hall, MSU, East Lansing, MI 48824.

You indicate your voluntary participation by completing and returning the survey on
questionpro.com.

~159-

Survey Instrument - Solicitation Email

DEVELOPMENT OF METHODOLOGY TO UNDERSTAND CAUSE AND EFFECT
RELATIONSHIP BETWEEN PRODUCTION DELAY FACTORS AND CONSTRUCTION
SITE PERFORMANCE

The Michigan State University Center for Construction Project Performance Assessment and
Improvement is conducting a research project to explore the relationship between production
delay factors and construction site performance. The research will study the current efforts in
construction industry to maintain reliable workflow on construction work site and identify the
production factors that have an effect on construction performance. The research will develop a
method to study the cause and effect relationship that will assist construction professionals to
build their company specific models to increase their understanding of their own company
construction operations. Funding is being provided by the Construction Project Performance
Assessment and Improvement, School of Planning Design and Construction, Michigan State
University. As part of the research we are surveying professionals who are involved with
construction projects as a part of the management team. As experienced industry participants,
their insight into the onsite construction production factors will be very useful to attaining the
aims of this research. The minimum qualification requirements for a participant are:

I. Must have at least 1 year or 1 project experience on construction project functioning as a
member/leader of the management team and used Last Planner system.

We, therefore, request you to forward this survey to the list serve that matches our requirements
within your company, asking for participation in the survey. As a participant in this research, they
will be asked a series of closed ended questions relating to construction onsite management
through an online survey. The participation is voluntary and anyone may choose to terminate
their involvement in this study at any time during this project. Please note that the research is
purely for academic purposes and survey results will not be shared with the company or the
participants. A participant if uncomfortable at any time during the survey, may terminate and
withdraw from the survey. They may refuse to answer any particular interview question. Their
privacy will be protected to the maximum extent allowable by law.

The estimated time to complete this survey is approximately 30-40 minutes. If you have any
questions about this project, you may contact Dr. Tariq Abdelhamid, School of Planning, Design
and Construction, Michigan State University at (517) 432-6188, or Samarth Jain, School of
Planning, Design and Construction, Michigan State University at (517) 203-9010. If you have any
questions or concerns about your role and rights as a research participant or would like to obtain
information or offer input, or would like to register a complaint about this research study, you
may contact, anonymously if you wish, Michigan State University Human Research Protection
Program at 517-355-2180, FAX 517-432-4503, or e-mail: irb@msu.edu, or regular mail at: 207
Olds Hall, MSU, East Lansing, MI 48824

-l60-

Survey Instrument

The survey instrument is a web based questionnaire reproduced in word format for
publishing purposes below.

Please enter your job position/title

 

 

 

 

Please enter no. of years of your working experience

 

 

 

 

Please enter your experience in Construction Management team/role

 

 

 

 

Enter an estimated no. of projects you have worked in management role

 

 

 

 

What is the average duration of projects (in months) that you have worked on?

 

 

 

 

In the following sections, you will be asked questions from seven categories. The first six
categories represent Production delay factors that come into action during the progress of

~161-

work namely:

. Completion of Prerequisite Work

. Availability of Directives (Information)

. Availability of Resources: Equipment, Labor, Material and Space.
. Propensity to Reduce Waste

. Propensity to Reduce Variation

. Propensity to Reduce Burden

QUIAUJNH

Each category has a list of questions presented as statements of fact representing an
aspect of production planning. A scale is provided next to each question. You are
requested to select the best value according to your experience. The last category i.e.
Work Flow Reliability, contains questions on work performance as result of the six
production planning categories encountered during work. Your answers from the most
recent project work week will be preferred, otherwise please fill responses based on your
cumulative experience.

Completion of Pre Requisite Work (PRE)

PRE l . Prior work is complete with desired quality.

 

l 2 3
( 10-100% Complete) Cl Cl Cl Cl Cl C] E] C1 Cl

.5
LII
O\
\l
00
\O
S

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PRE2. The pre-requisite work has been verified to meet current work needs (dimensions,
locations, etc)

 

l 2 3 4 5 6 7
( 10-100% Agree) D D Cl Cl Cl Cl Cl Cl Cl D

00
SO
0

 

 

 

 

 

 

 

 

 

 

 

 

 

PRE3. All regulatory inspections needed were conducted and successfully passed.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10

( 10-100% Agree) D E] Cl 0 D 0 Cl C] D D
PRE4. The prerequisite work is officially handed over from prior trade.

1 2 3 4 5 6 7 8 9 10

L0 0-100% Agree) Cl Cl C1 C1 Cl E] Cl C! D D

 

 

 

 

 

 

 

 

 

 

 

 

~162-

PRES . The weather forecast is favorable for performing work.

 

I

2

3

4

5

10

 

 

( 10-100% Agree)

 

Cl

 

Cl

 

F1

 

fl

 

,__.
‘ l
l |

 

 

 

 

Cl

 

LJ

 

PRE6. Access to the work area (permits, etc) has been obtained and cleared with respective party.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10
( 10-100% Agree) L '. _! L? L! L‘ ;.l I] '_ f_. .
PRE7. All coordination issues with following trade have been addressed by the project team.
I 2 3 4 5 6 7 8 9 10
( 10-100% Agree) ’7 i " I ' S I] [l fl [7 fl
PRE8. Overall, the prerequisite work is complete.
I 2 3 4 5 6 7 8 9 10
( 10-100% Complete) ‘ ' LI Ll ll 3 D L 1-] U L)
Availability of Directives (Information) (DIR)
DIR] . 'Ihe D&E documents for current work are acceptable.
1 2 3 4 5 6 7 8 9 10
(10- 1 00% Agree) L: U 3' [3 CI 13 a U D
DIR2. The D&E documents for current work are available.
1 2 3 4 5 6 7 8 9 10
( 10-100% Agree) Ll D t] L'_: L1 13 E] L1 LI 9
DIR3. All RFIs for current work are addressed satisfactorily.
1 2 3 4 5 6 8 9 10
( 10-100% Agree) E. l L“ C.) C} 1; Cl 3- D {j 3
DIR4. The owner is agreeable to any design/scope changes, if any.
r‘"? 1 2 3 4 5 6 7 8 9 10
( 1 0-100% Agree) L LI 5' [I I: D L? LJ Ll [j

 

 

 

 

 

 

 

 

 

 

 

 

—l63-

 

 

 

 

 

 

DIRS. The standard of work performance is available.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 4’7 l}? ‘ r; ‘ I [I [I T! [I [l
DIR6. The standard of work performance is clearly expressed.
l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) El Tl CT W [3 Cl I: [3 E D
DIR7. The submittals are approved.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) *1 l ' L} L L‘ L‘ L L; L1 L1
DIR8. The submittals are available.
I 2 3 4 5 6 7 8 9 10
(IO-100% Available) '7 t7 I: [l E P. [l 1? El. {3
DIR9. Final design instructions have been confirmed by architect/engineer to avoid change.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agee) l". l? U If? P. [I E‘ [I ll
DIR10. All lien waivers are in order.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) I? 1‘1 I] [1 D I? D D D
DIRI 1. All permits are available.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) -‘" ll [7 D " I l l ‘ It I]
DIR12. All lien waivers are submitted.
1 2 3 4 5 6 7 8 9 10
(IO-100% Complete) [1 '3 *"7 Cl C!- ‘fl D D '3 l]-

 

 

 

 

 

 

 

 

 

 

 

 

-l64-

 

 

 

 

 

 

 

 

DIR13. Overall, all the directives are in order.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I 2 3 4 6 8 9 IO
(IO-100% Agree) " M e__' l . U l. " LI 1;
Availability of Resources: Equipment, Labor, Material and Space (RES)
RESl. The material is/will be delivered on time.
1 2 3 4 6 8 9 10
(IO-100% Agree) L1 Li :1 LJ 1 L7 L1 Li
RES2. The material specifications match to contract/submittal specs.
I 2 3 4 6 8 9 10
(IO-100% Agree) 7| {‘3 '7 l7 f? T l ’" [T-
RES3. Space is available for material lay-downs.
l 2 3 4 6 8 9 10
(IO-100% Available) [1 I ' [3 L [I L L? J
RES4. Space is available for material transport.
1 2 3 4 6 8 9 10
(IO-100% Available) I ‘ ' ‘ ‘ 1 I] [.1 Q; ‘ ‘
RESS. The labor is informed of all work assignments.
l 2 3 4 6 8 9 10
(IO-100% Agree) L l; l :1 LJ 8 L: a 5
RES6. The labor needed is available.
I 2 3 4 6 8 9 10
(IO-100% Available) I? l l i ‘ If] 7 Vi l1
RES7. The tools to be used are in good working condition.
1 2 3 4 6 8 9 10
(IO-100% Agree) L L Li L} L7 ‘ ' i _1

 

 

 

 

 

 

 

 

 

 

 

 

-l65-

 

 

 

 

 

 

 

 

 

RESS. Right tools are available in enough quantity for crews to work with.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 2 3 4 5 6 7 8 9 10
(IO-100% Available) " :J L. l] L C l L1 D L1 :1
RES9. The equipment to be used is in good working condition.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 1 l V‘ ' i 7‘ l fl f“: 1'3 71 fl
RESIO. Right equipment is available in the desired quantity.
1 2 3 4 5 6 7 8 9 IO
(IO-100% Available) L. 7 L1 Li L [I 1 fl L] l’ l
RESI 1. Overall, all the required resources are available.
1 2 3 4 5 6 7 8 9 10
(IO-100% Available) ._| L J .I L L. L‘ ‘? ' Ll ' L
Propensity to Reduce Waste (WTE)
WTE]. The amount of work is sized appropriately to avoid overproduction.
l 2 3 4 5 6 7 8 9 10
( 10-100% Agree) L ‘ -' 7 ‘

 

 

 

 

 

LI

 

l

 

l l

 

 

 

 

 

WTE2. Transportation of raw/processed materials requires single handling (as opposed to more

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

than once).
2 3 4 6 7 8 10
(10- l 00% Agree) ”T 7 T, f} r. ‘5, f1 ;- .1 5| ,_-.
WTE3. Transportation of raw/processed materials is over short distances.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) CI .7 ' ‘ “j '1 : f. q
WTE4.No amount of rework/correction is required for completed work.
1 2 3 4 5 6 8

(IO-100% Agree)

 

 

 

 

 

 

 

 

 

 

 

-l66-

 

 

 

 

 

 

 

 

WTES. No over-processing (unnecessary finishing) is performed for completed work.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 2 3 4 5 6 7 8 10
(IO-100% Agree) L‘ L E [1 I] L D L I: E
WTE6. Inventory is delivered just-in-time to the production unit.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) ll [1 m '71 fl " n m n
WTE7. The production unit makes no unnecessary movements to complete the work.
1 2 3 4 5 6 7 8 9 10
( 10-100% Agree) I] I: L D C: L '3 E] L D
WTE8. The production unit is always working to complete the work.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) L’ L1 D U U L1 L1 ll l] [1
WTE9. The production crew is consulted on the best way to perform the work.
I 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 17' 27 H II. H H [I U [1
WTE10. The production crew is consulted on the safest way to perform the work.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) L L ’7 D '1 I] L“ “1 Cl I]
WTE] 1. Overall, sufficient steps have been taken to prevent generation of waste.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 11 U L‘ L 1 LI L LI L' U
Propensity to Reduce Variation (VAR)
VARI. There was no absenteeism in the workforce.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) L L L" Ll D L L? 1.: L‘ L‘

 

 

 

 

 

 

 

 

 

 

 

 

-l67-

 

 

 

 

 

 

 

 

VAR2. The production crews operate with reserve capacity.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 4 5 6 7 8 9 10
(10- l 00% Agree) L .3 L L‘ [3 Cl L C1 C1 D
VAR3. The learning curve effect is realizable for the crews.
I 2 3 4 5 6 7 9 10
(10- l 00% Agree) ‘ T! 71- 17 {"1 Lil Fl _. C1 17'
VAR4. Recovery plans from occupational accidents are standard practice.
I 2 3 4 5 6 8 9 10
( 10-100% Agree) L D .3 [.1 El D L D L L
VARS. The production rates of the crew are consistent over time (match Takt).
l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) L1 11 u u Ll L1 L1 L D Ll
VAR6. The production plan has workable backlog in case of interruptions to current
assignment(s).
I 2 3 4 5 6 7 8 9 10
(IO-100% Agree) [7 :1 [—1 T1 F] D [1 [1 [l
VAR7. The production plan has time buffers to prevent interruptions to project completion.
1 2 3 4 S 6 7 8 9 10
( 10-100% Agree) L C“ Li [I [I E] L L‘ f] L]

 

 

 

 

 

 

 

 

 

 

 

 

Propensity to Reduce Burden (BUR)

BUR]. The work environment is free of potential environmental hazards (air-bome pathogens,

dust, chemical agents, ultra-violet light, and ionizing radiation).

 

l

2

3

4

5

\l

10

 

( 10-100% Agree)

 

 

Ll

 

L!

 

[I
__l

 

 

D

 

 

[:1

 

 

 

Ll

 

-l68-

 

 

 

 

 

 

 

BUR2. The work environment is comfortable to function in (temperature and humidity, lighting,

sun exposure, noise levels).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) gr , l . l 'l L7 13 .l L. L‘ L'
BUR3. The work environment is not congested.
l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) " ‘ ‘ l .' ’ l L: L .' '__l ; Ll
BUR4. Overtime is not needed on our jobs.
I 2 3 4 5 6 7 8 9 10
(IO-100% Agree) ' ‘ i ' 7 L? 'J L L“ I? L] C
BURS. The work is adequately paced to avoid physical fatigue.
1 2 3 4 5 6 7 8 9 10
( 10-100% Agree) ' " I ' r l ' 11;! '_ .‘ ' I "
BUR6. The workers are equipped with PPEs, as necessary.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) Ll fl ': I: f7 I? L1 [1 1’1 [1
BUR7. The workers are not allowed to engage work involving heavy muscular loads.
1 2 3 4 5 6 7 8 9 IO
(IO-[00% Agree) " Ll " U ll ll ll ll [1 ll
BUR8. The workers have access to material-lifting equipment.
I 2 3 4 5 6 7 8 9 10
(10- I 00% Agree) IT } Fl fl D I? U H 7 fl D
BUR9. A Job Safety Analysis is performed.
1 2 5 10

 

E-100% Agree)

 

 

 

 

 

 

 

 

 

 

 

-l69-

 

 

 

 

 

 

 

 

BURIO. Workers are trained on proper lifting techniques.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 8 9 10
(IO-100% Agree) ‘71 T [1 If]. i] I] I? . l U I]
BUR] 1. Rest cycles are built into the work method.

I 2 3 4 5 7 8 9 10

(IO-100% Agree) ' ' " "‘ 7' 7 7 "1 ’—. f7
BUR12. Workers receive sufficient information regarding the process flow and output.

I 2 3 4 5 6 7 8 9 10

T ‘ I I I _ M :1 [[3

 

(IO-100% Agree)

 

 

 

 

_J

 

 

 

 

 

 

 

BUR13. The rate of information does not exceed the mental capacity of the worker to process.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) _. J L _l l; L l; 1 1 LI L‘
BUR14. Identical or very similar signals don’t occur for a long time.
I 2 3 4 5 6 7 8 9 10
(IO-100% Agree) "1 T} 7" r 2 fl 5”" Fl “
BURIS. Adequate time is allowed for decisions and resulting actions in the normal
circumstances.
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) " 5‘ "' ’"I F' ‘ ' a w
BUR16. Adequate time is allowed for decisions and resulting actions in emergencies.
l 2 3 4 5 6 7 8 9 10
( 10-100% Agree) ' 1 Ta 7 I] C F" T? Tl
EUR] 7. Hand tools used are the correct ones for the task.
4 7

l

2

3

10

 

EIOO% Agree)

 

 

 

 

 

 

 

 

 

 

 

-170-

 

 

 

 

 

 

 

 

BUR18. Hand tools used are adequately maintained.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) L Ll ”- Li L? C." [I l } L L“
BUR19. All safety signs and visuals are correctly located.
I 2 3 4 5 6 7 8 9 10
( 10-100% Agree) Ll Ll i Ll L '..j Ll L! _l U
BUR20. Overall, the workers do not feel any kind of burden.
l 2 3 4 5 6 7 8 9 10
(IO-100% Agree) Tl i? Cl :1 [l [1 L1 El D I]
Reliable Work Flow Achieved (WF R)
WFR]. The PPC is high; 90% or higher
I 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 0 D D D Cl E] El Cl Cl Cl
WFR2. We are headed to a timely completion of the project
1 2 3 4 5 6 7 8 9 10
(IO-100% Agree) 0 D D 0 El D CI D D D

 

 

 

 

 

 

-l7l-

 

 

 

 

 

 

 

 

 

 

 

EQS, A STRUCTURAL EQUATION PROGRAM

APPENDIX B

Output File from EQS for Framework Demonstration Example

MULTIVARIATE SOFTWARE,

INC.
COPYRIGHT BY P.M. BENTLER VERSION 6.1 (C) 1985 -
2009 (B97).
PROGRAM CONTROL INFORMATION
1 /TITLE
2 Model built by EQS 6 for Windows
3 /SPECIFICATIONS
4 DATA:'c:\eqs6l\support\sam\thesis 400 samples.ess';
S VARIABLES=72; CASES=400;
6 METHOD=ML; ANALYSIS=COVARIANCE; MATRIX=RAW;
7 /LABELS
8 V1=PRE1; V2=PRE2; V3=PRE3; V4=PRE4; V5=PRE5;
9 V6=PRE6; V7=PRE7; V8=PRE8; V9=DIR1; V10=DIR2;
10 V11=DIR3; V12=DIR4; V13=DIR5; V14=DIR6; V15=DIR7;
11 V16=DIR8; V17=DIR9; V18=DIR10; V19=DIR11; V20=DIR12;
12 V21=DIR13; V22=RESl; V23=RES2; V24=RES3; V25=RES4;
13 V26=RESS; V27=RES6; V28=RES7; V29=RESB; V30=RES9;
14 V31=RESlO; V32=RESll; V33=WTE1; V34=WTE2; V35=WTE3;
15 V36=WTE4; V37=WTE5; V38=WTE6; V39=WTE7; V40=WTE8;
16 V41=WTE9; V42=WTE10; V43=WTE11; V44=VAR1; V45=VAR2;
l7 V46=VAR3; V47=VAR4; V48=VAR5; V49=VAR6; V50=VAR7;
18 V51=BUR1; V52=BUR2; V53=BUR3; V54=BUR4; V55=BUR5;
19 V56=BUR6; V57=BUR7; V58=BUR8; V59=BUR9; V60=BUR10;
20 V61=BUR11; V62=BUR12; V63=BUR13; V64=BUR14; V65=BUR15;
21 V66=BUR16; V67=BUR17; V68=BUR18; V69=BUR19; V70=BUR20;
22 V71=WFR1; V72=WFR2;
23 /EQUATIONS
24 V8 = 1F1 + E8;
25 V21 = 1F2 + E21;
26 V32 = 1F3 + E32;
27 V43 = 1F4 + E43;
28 V48 = 1F5 + E48;
29 V70 = 1F6 + E70;
30 V71 = 1F7 + E71;
31 V72 = *F7 + E72;
32 F1 = *Vl + *V2 + *V3 + *V4 + *V5 + *V6 + *V7
33 ;
34 F2 = *V9 + *V10 + *Vll + *V12 + *V13 + *V14 + *V15
35 + *V16 + *V17 + *V18 + *Vl9 + *V20 ;
:36 F3 = *V22 + *V23 + *V24 + *V25 + *V26 + *V27 + *V28
:37 + *V29 + *V30 + *V31 ;
:38 F4 = *V33 + *V34 + *V35 + *V36 + *V37 + *V38 + *V39
39 + *V40 + *V41 + *V42 ,
4CD F5 = *V44 + *V45 + *V46 + *V47 + *V49 + *VSO ;
411 F6 = *V51 + *V52 + *V53 + *V54 + *V55 + *V56 + *V57
422 + *V58 + *V59 + *V6O + *V6l + *V62 + *V63 + *V64
4:3 + *V65 + *V66 + *V6? + *V68 + *V69 ;

-l72-

44 F7 : *Fl + *F2 + *F3 + *F4 + *F5 + *F6 + D7;
45

46 /VARIANCES

47 Vl = *;

48 V2 = *;
49 V3 = *;
50 V4 = *;
51 V5 = *;
52 V6 = *;
53 V7 = *;
54 V9 = *;
55 V10 = *;
56 V11 = *;
57 V12 = *;
58 V13 = *;
59 V14 = *;
60 V15 = *;
61 V16 = *;
62 V17 = *;
63 V18 = *;
64 V19 = *;
65 V20 = *;
66 V22 - *;
67 V23 = *;
68 V24 = *;
69 V25 = *;
70 V26 = *;
71 V27 = *;
72 V28 = *;
73 V29 = *;
74 V30 = *;
75 V31 = *;
76 V33 = *;
77 V34 = *;
78 V35 = *;
79 V36 = *;
80 V37 = *;
81 V38 = *;
82 V39 = *;
83 V40 = *;
84 V41 = *;
85 V42 = *;
86 V44 = *;
87 V45 = *;
88 V46 = *;
89 V47 = *;
90 V49 = *;
91 V50 = *;
92 V51 = *;
93 V52 = *;
94 V53 = *;
95 V54 = *;
96 V55 = *;
97 V56 = *;
98 V57 = *;
99 V58 = *;
100 V59 = *;

-173-

101 V60 = *;
102 V61 = *;
103 V62 = *;
104 V63 = *;
105 V64 = *;
106 V65 = *;
107 V66 = *;
108 V67 = *;
109 V68 = *;
110 V69 = *;
111 E8 = *;
112 E21 = *;
113 E32 *;
114 E43 = *;
115 E48 = *;
116 E70 = *;
117 E71 = *;
118 E72 = *;
119 D7 = *;
120 /COVARIANCES
121 /PRINT
122 EIS;

123 FIT=ALL;

124 TABLE=EQUATION;

125 /END

125 RECORDS OF INPUT MODEL FILE WERE READ

DATA IS READ FROM c:\eq361\support\sam\thesis 400 samples.ess
THERE ARE 72 VARIABLES AND 400 CASES
IT IS A RAW DATA ESS FILE

SAMPLE STATISTICS BASED ON COMPLETE CASES

UNIVARIATE STATISTICS

VARIABLE PRE1 PRE2 PRE3 PRE4 PRES
V1 V2 V3 V4 V5
MEAN 5.6050 5.4225 5.5225 5.5125
5.4275
SKEWNESS (G1) -0.0559 0.0449 0.0847 -0.0354
0.0456
KURTOSIS (G2) -1.2527 -1.1995 -1.l437 -1.2394
1.2815
STANDARD DEV. 2.9590 2.7962 2.7980 2.8522
2.9693
VARIABLE PRE6 PRE7 PRE8 DIR1 DIR2
V6 V7 V8 V9 V10

-174-

MEAN
.4200

SKEWNESS
.0592

KURTOSIS
.2823

STANDARD

.9750

VARIABLE

MEAN

.7275

SKEWNESS
.0941

KURTOSIS
.2906

STANDARD

.9504

VARIABLE

MEAN

.4900

SKEWNESS
.0178

KURTOSIS
.2484

STANDARD

.8522

VARIABLE

MEAN

.5225

SKEWNESS
.0614

KURTOSIS
.2615

DEV.

DEV.

(G1)

(62)

DEV.

(G1)

(G2)

5.4425

0.0492

-1.1765

2.8190

DIR3

V11

5.8775

-0.1820

-1.1785

2.8377

DIR8

V16

5.4250

0.0351

-1.3000

2.9455

DIR13

V21

5.4450

—0.0198

-1.2661

5.6250

-0.0470

-1.2906

2.8513

DIR4

V12

5.4400

-0.0620

-1.3308

2.9390

DIR9

V17

5.4150

0.0183

-1.3349

3.0017

RESl

V22

5.3325

0.0655

-1.2017

-175-

5.6400

-0.1050

-1.3034

2.9097

DIRS

V13

5.6875

-0.0578

-1.1933

2.7927

DIR10

V18

5.6125

-0.1208

~1.2143

2.8693

RESZ

V23

5.2775

0.1023

-1.2449

5.7425

-0.1880

-1.0715

2.7587

DIR6

V14

5.4575

0.0455

-1.2002

2.8192

DIR11

V19

5.6625

-0.0616

-1.1993

2.8485

RES3

V24

5.7375

-0.0916

-1.2384

DIR7
V15

DIR12
V20

RES4
V25

STANDARD
.9693

VARIABLE

MEAN

.7450

SKEWNESS
.1148

KURTOSIS
.2390

STANDARD

.9165

VARIABLE

MEAN

.4075

SKEWNESS
.0790

KURTOSIS
.3481

STANDARD

.0569

VARIABLE

MEAN

.3825

SKEWNESS
.0321

KURTOSIS
.1150

STANDARD

.7930

VARIABLE

MEAN
.4200

DEV.

(G1)

(G2)

DEV.

(G1)

(G2)

DEV.

(G1)

(G2)

DEV.

2.9255

RESS

V26

5.4175

0.0118

~1.2422

2.8729

RESlO

V31

5.4925

-0.0027

-l.2519

2.8408

WTE4

V36

5.4700

0.0182

-l.2081

2.8371

WTE9
V41

5.4675

2.8464

RES6

V27

5.4875

-0.0590

-1.2011

2.9456

RESll

V32

5.6300

-0.1134

-1.2217

2.8589

WTES

V37

5.5625

-0.0197

-1.2311

2.9012

WTE10
V42

5.5825

-l76-

2.8977

RES7

V28

5.5375

-0.0699

—1.2200

2.8458

WTEl

V33

5.6525

-0.0903

-1.2088

2.8136

WTE6

V38

5.6300

-0.0549

-1.2067

2.8721

WTE11
V43

5.7350

2.8791

RESB

V29

5.6200

-0.0463

-1.1774

2.8847

WTE2

V34

5.5050

-0.0478

-1.2060

2.8443

WTE7

V39

5.3025

0.1134

-1.2356

2.9167

VARl
V44

5.8675

RES9
V30

WTE3
V35

WTE8
V40

VAR2
V45

SKEWNESS
.0426

KURTOSIS
.2305

STANDARD

.9008

VARIABLE

MEAN

.4925

SKEWNESS
.0630

KURTOSIS
.2022

STANDARD

.8408

VARIABLE

MEAN

.2500

SKEWNESS
.0669

KURTOSIS
.1011

STANDARD

.7869

VARIABLE

MEAN

.2775

SKEWNESS
.1121

KURTOSIS
.2973

STANDARD
.8620

(G1)

(G2)

DEV.

DEV.

(G1)

(G2)

DEV.

DEV.

0.0451

-1.2126

2.8853

VAR3

V46

5.4725

-0.0363

-1.2349

2.8949

BURl

V51

5.6800

—0.1307

-1.2364

2.8597

BUR6

V56

5.6700

-0.1095

—1.1905

2.8144

-0.0571

-1.2442

2.8493

VAR4

V47

5.3100

0.0642

-1.2882

2.9437

BUR2

V52

5.6500

-0.0547

-1.2203

2.8789

BUR7

V57

5.6425

-0.0880

~1.2582

2.9345

-177-

-0.0713

-1.2214

2.8724

VARS

V48

5.3175

0.0883

-1.3367

2.9974

BUR3

V53

5.6375

-0.l472

-l.1913

2.8735

BUR8

V58

5.3900

0.1009

~1.2133

2.8448

-0.1781

-1.2646

2.9062

VAR6

V49

5.7825

-0.0516

-1.2658

2.9157

BUR4

V54

5.4875

-0.0175

-1.2685

2.9096

BUR9

V59

5.5100

-0.0049

-1.l897

2.8275

 

VAR7
V50

BURS
V55

BUR10
V60

VARIABLE

MEAN

.4050

SKEWNESS
.0516

KURTOSIS
.1508

STANDARD

.7677

VARIABLE

MEAN

.3975

SKEWNESS
.0902

KURTOSIS
.2226

STANDARD

.8442

VARIABLE

MEAN

SKEWNESS

KURTOSIS

STANDARD

MARDIA'S COEFFICIENT

DEV.

DEV.

(G1)

(G2)

DEV.

BURll
V61

5.4850

-0.0554

-1.2191

2.8715

BUR16
V66

5.4225

0.0102

—1.1897

2.7908

WFRI
V71

5.4150

0.0436

-1.2777

2.8746

NORMALIZED ESTIMATE =

(G2,P)

BUR12
V62

5.1225

0.1228

-1.1768

2.8492

BUR17
V67

5.0925

0.1576

-1.2600

2.8800

WFR2
V72

5.5275

-0.0189

-1.2742

2.8600

BUR13

V63

5.4625

0.0192

-1.2231

2.7997

BUR18

V68

5.5075

0.0323

-1.2306

2.8628

MULTIVARIATE KURTOSIS

= -86.9022
-8.4185

BUR14

V64

5.4750

0.0176

-l.2234

2.8740

BUR19

V69

5.6125

-0.0419

-1.1769

2.8315

ELLIPTICAL THEORY KURTOSIS ESTIMATES

-l78-

BUR15
V65

BUR20
V70

MARDIA-BASED KAPPA =

-O.4100

-0.0163 MEAN SCALED UNIVARIATE KURTOSIS

MARDIA-BASED KAPPA IS USED IN COMPUTATION. KAPPA:

CASE NUMBERS WITH LARGEST CONTRIBUTION TO NORMALIZED MULTIVARIATE

KURTOSIS:

CASE NUMBER
229

ESTIMATE
296.4353

COVARIANCE MATRIX TO BE ANALYZED:

239.9667

VARIABLES)
BASED ON 400 CASES.
PRE1
PRES
V1

V5

PRE1 V1 8.756

PRE2 V2 -O.372

PRE3 V3 0.079

PRE4 V4 -0.363

PRES V5 1.227
8.817

PRE6 V6 0.050
0.763

PRE7 V7 -0.061
0.366

PRE8 V8 0.211
0.781

DIR1 V9 0.016
0.353

DIR2 V10 -0.380
0.765

DIR3 V11 -0.226
0.657

DIR4 V12 0.560
0.153

DIRS V13 -0.124
0.131

DIR6 V14 -0.S66
0.326

DIR7 V15 1.286
0.254

DIR8 V16 —0.225
0.355

DIR9 V17 -0.204
0.491

13

286.154

PR

2

E2

V2

-0.

-0.

-l79-

.819

620

.126

592

.716

.800

.516

.139

.506

.220

.533

.263

.345

.103

.336

.464

288.

72 VARIABLES

16

7188

231.1765.

-0.0163

133

(SELECTED FROM

PRE3

V3

.829

.131

.014

.708

.773

.661

.091

.716

.384

.025

.236

.367

.223

.025

.091

PRE4

V4

.135
.289

.204

.291

.512

.260

.501

.040

.444

.724

.001

.323

.103

.456

 

DIR10
.754
DIR11
.212
DIR12
.759
DIR13
.293
RESI
.336
RES2
.518
RESB
.289
RES4
.202
RESS
.280
RES6
.161
RES7
.333
RESB
.261
RES9
.037
RESIO
.838
RESll
.240
WTEl
.056
WTE2
.452
WTE3
.089
WTE4
.264
WTES
.414
WTE6
.379
WTE7
.241
WTE8
.162
WTE9
.233
WTE10
.408
WTE11
.429
VARl
.530
VAR2
.829

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

.117

.067

.530

.011

.337

.909

.670

.219

.093

.652

.163

.103

.729

.032

.510

.639

.216

.001

.100

.175

.204

.304

.302

.141

.744

.599

.273

.097

-l80-

.194

.521

.265

.467

.216

.035

.415

.197

.112

.044

.597

.038

.386

.356

.480

.034

.602

.509

.047

.145

.134

.148

.357

.035

.354

.371

.247

.890

.216

.949

.623

.712

.242

.393

.304

.070

.110

.710

.024

.194

.753

.007

.007

.220

.057

.561

.667

.036

.758

.190

.333

.359

.049

.039

.043

.141

.565

.198

.089

.176

.148

.311

.437

.090

.242

.218

.032

.439

.221

.043

.669

.253

.708

.430

.540

.191

.110

.644

.948

.206

.267

.814

.040

.666

VAR3
.328
VAR4
.326
VARS
.290
VAR6
.456
VAR7
.476
BURl
.270
BUR2
.844
BUR3
.514
BUR4
.675
BURS
.425
BUR6
.325
BUR7
.103
BUR8
.234
BUR9
.383
BUR10
.154
BURll
.349
BUR12
.128
BUR13
.970
BUR14
.757
BUR15
.674
BUR16
.314
BUR17
.533
BUR18
.446
BUR19
.177
BUR20
.020
WFRl
.413
WFR2
.818

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

.106

.278

.101

.044

.186

.708

.611

.758

.153

.254

.040

.046

.475

.001

.975

.420

.648

.198

.103

.263

.092

.237

.126

.979

.549

.337

.525

-l8l-

.502

.322

.155

.140

.254

.168

.176

.056

.522

.185

.301

.477

.682

.483

.095

.326

.146

.113

.487

.633

.102

.028

.141

.094

.243

.977

.383

.155

.036

.227

.307

.509

.686

.000

.657

.343

.268

.163

.570

.981

.052

.060

.334

.893

.104

.305

.252

.603

.210

.812

.727

.040

.320

.640

.335

.475

.491

.307

.201

.240

.127

.047

.601

.346

.512

.220

.281

.126

.316

.402

.757

.003

.665

.073

.412

.085

.516

.119

.390

.008

.554

DIR2
V10
PRE6
PRE7
PRE8
DIR1
DIR2
8.851
DIR3
0.016
DIR4
1.153
DIRS
0.264
DIR6
0.599
DIR7
0.394
DIR8
0.535
DIR9
0.201
DIR10
0.153
DIR11
0.027
DIR12
0.769
DIR13
0.704
RESl
0.196
RESZ
0.008
RES3
0.223
RES4
0.265
RESS
0.682
RES6
0.769
RES7
0.127
RESS
1.259
RES9
0.092
RESlO
0.167
RESll
0.313
WTEl
0.071

V6

V7

V8

V9

V10

V11

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

PRE6

V6

.947

.094

.706

.397

.300

.465

.696

.161

.276

.147

.279

.252

.585

.118

.041

.130

.351

.684

.059

.294

.356

.417

.037

.355

.305

.398

.004

.608

PRE7

V7

-182-

.130

.188

.179

.173

.141

.830

.819

.242

.411

.466

.210

.170

.502

.623

.353

.579

.329

.081

.716

.199

.329

.357

.855

.259

.421

.177

.000

PRE8

V8

.467

.311

.508

.026

.445

.338

.315

.187

.922

.051

.617

.081

.094

.078

.306

.413

.231

.083

.461

.522

.470

.561

.764

.511

.403

.276

DIR1

V9

.610
.106

.320

.166

.373

.203

.120

.022

.528

.579

.120

.099

.514

.536

.778

.000

.203

.140

.082

.149

.324

.057

.368

.067

.259

WTE2
.235
WTE3
.598
WTE4
.461
WTES
.668
WTE6
.130
WTE7
.265
WTE8
.044
WTE9
.560
WTE10
.087
WTE11
.177
VARl
.202
VAR2
.357
VAR3
.568
VAR4
.772
VARS
.270
VAR6
.615
VAR7
.434
BURl
.009
BUR2
.092
BUR3
.639
BUR4
.085
BURS
.108
BUR6
.470
BUR7
.313
BUR8
.149
BUR9
.125
BUR10
.267
BUR11
.635

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

.515

.081

.537

.260

.440

.164

.018

.115

.240

.368

.508

.029

.046

.005

.146

.497

.173

.076

.421

.143

.126

.250

.645

.400

.017

.434

.336

.240

-183-

.543

.478

.695

.470

.177

.024

.690

.709

.129

.017

.975

.178

.602

.427

.470

.259

.072

.266

.342

.018

.534

.410

.287

.533

.352

.432

.423

.555

.369

.152

.282

.298

.663

.189

.008

.345

.716

.236

.053

.064

.030

.468

.413

.323

.261

.697

.162

.612

.110

.223

.084

.149

.632

.234

.178

.098

.236

.130

.265

.075

.361

.619

.064

.040

.070

.114

.134

.187

.284

.722

.413

.099

.059

.729

.822

.134

.348

.073

.253

.281

.351

.217

.014

.025

BUR12
0.161

BUR13
0.393

BUR14
0.107

BUR15
0.381

BUR16
0.002

BUR17
0.189

BUR18
0.315

BUR19
0.862

BUR20
0.051

WFR1
0.262
WFR2

0.603

DIR7

V15
DIR3
DIR4
DIRS
DIR6
DIR7
8.705
DIR8
0.914
DIR9
0.476
DIR10
0.360
DIR11
0.449
DIR12
0.259
DIR13
1.024
RESl
0.379
RESZ
0.126
RES3
0.300
RES4
0.095

RESS
0.157

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V11

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

0.121
-0.263

0.163

0.229
-0.621
—0.352
-0.634
-0.232
-0.154
-0.279

0.405

DIR3
V11
8.053
0.187
0.207
0.247
0.177
-0.028
-0.693
-0.990
0.162
-0.073
0.137
0.439
-0.257
0.071

0.801

0.004

0.016

0.194

0.903

0.496

-0.748

0.561

0.163

0.117

-0.460

0.271

0.193

DIR4

V12

8.638

0.168

-0.169

0.248

0.221

0.293

-0.260

-0.257

-0.138

-0.429

0.021

0.374

0.311

-0.554

-1.104

-184-

-0.487

-0.031

-0.092

-0.212

0.586

0.542

-0.709

0.249

0.158

0.100

4.769

DIRS

V13

7.799

-0.428

-0.098

-0.932

-0.572

-0.081

—0.291

-0.162

-0.196

-0.156

0.593

0.394

-0.027

-0.007

-0.314

0.159

0.245

0.598

0.412

0.054

0.048

0.186

0.491

-0.459

0.144

DIR6

V14

7.948
—0.035

—0.420

-0.321

-0.213

-0.610

0.264

0.458

0.151

-0.531

0.386

-0.433

0.402

RES6
.005
RES7
.287
RESB
.079
RES9
.161
RESlO
.092
RESll
.577
WTE1
.393
WTE2
.183
WTE3
.565
WTE4
.595
WTES
.083
WTE6
.432
WTE7
.466
WTE8
.217
WTE9
.529
WTE10
.044
WTE11
.101
VARl
.377
VAR2
.011
VAR3
.455
VAR4
.078
VARS
.465
VAR6
.415
VAR7
.199
BUR1
.601
BUR2
.121
BUR3
.061
BUR4
.577

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

.331

.252

.217

.455

.403

.967

.689

.580

.556

.436

.300

.243

.968

.162

.323

.264

.298

.039

.465

.341

.314

.141

.041

.261

.106

.368

.647

.567

-185-

.273

.425

.321

.111

.082

.681

.208

.052

.324

.023

.303

.173

.888

.177

.585

.061

.180

.550

.018

.230

.362

.281

.425

.357

.538

.761

.291

.517

.218

.145

.182

.404

.006

.461

.317

.208

.198

.308

.160

.771

.319

.206

.094

.334

.243

.638

.553

.414

.044

.338

.963

.026

.170

.145

.297

.055

.016

.693

.342

.405

.261

.522

.369

.059

.056

.221

.055

.491

.029

.206

.247

.417

.528

.053

.045

.232

.130

.689

.819

.025

.299

.381

.779

.219

BURS
0.073
BUR6
0.145
BUR7
0.163
BUR8
0.507
BUR9
0.079
BUR10
0.145
BUR11
0.098
BUR12
0.385
BUR13
0.575
BUR14
0.806
BUR15
0.025
BUR16
0.368
BUR17
0.060
BUR18
0.580
BUR19
1.308
BUR20
0.294
WFR1
0.101
WFR2
0.365

DIR12

V20
DIR8
DIR9
DIR10
DIR11
DIR12
8.135
DIR13
0.248
RESl
0.381
RESZ
0.084
RES3
0.721

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V16
V17
V18
V19
V20
V21
V22
V23

V24

-0.676

0.486

-0.049

-0.105

0.278

0.357

0.125

0.276

0.039

-0.007

-0.574

-0.324

-0.713

-0.286

0.115

0.129

-0.017

0.476

DIR8

V16
8.676
0.214
-0.674
~0.553
-0.768
-0.177
-0.179
-0.033

0.756

0.409

0.830

0.015

0.284

0.605

-0.163

0.342

-0.212

0.538

-0.034

0.746

0.087

0.323

-0.229

0.489

-0.689

0.601

-0.037

DIR9

V17

9.010

0.349

-0.669

0.435

-0.255

-0.281

0.093

0.157

-l86-

0.026

-0.061

0.284

0.283

-0.204

0.428

0.347

0.304

0.240

0.141

-0.267

0.318

-0.239

0.738

0.064

0.683

3.551

-0.644

DIR10

V18

8.233
-0.637
0.271
-0.739
-0.412
-0.055

-0.187

-0.012

-0.109

0.239

0.596

-0.933

0.186

-0.135

-0.141

0.006

-0.616

-0.171

-0.256

-0.265

-0.423

0.193

-0.298

-0.276

3.718

DIR11

V19

8.114
0.051

-0.802

-0.148

-0.109

0.307

RES4
.986
RESS
.033
RES6
.217
RES7
.147
RESB
.254
RES9
.421
RESlO
.097
RESll
.117
WTE1
.376
WTE2
.546
WTE3
.802
WTE4
.073
WTES
.207
WTE6
.302
WTE7
.934
WTE8
.323
WTE9
.926
WTE10
.722
WTE11
.163
VARl
.103
VAR2
.137
VAR3
.144
VAR4
.473
VARS
.067
VAR6
.395
VAR7
.039
BUR1
.016
BUR2
.317

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

.356

.088

.078

.059

.475

.389

.291

.116

.389

.449

.122

.190

.435

.471

.250

.025

.159

.714

.306

.307

.207

.265

.017

.596

.125

.043

.322

.540

-l87-

.083

.007

.085

.616

.163

.571

.324

.146

.205

.008

.611

.103

.740

.225

.486

.603

.292

.099

.301

.313

.454

.723

.139

.440

.644

.066

.008

.296

 

.273

.013

.292

.048

.097

.192

.344

.344

.077

.026

.531

.280

.103

.174

.308

.172

.886

.061

.115

.508

.133

.090

.266

.412

.126

.161

.124

.405

.011

.159

.428

.788

.028

.475

.287

.108

.005

.265

.401

.046

.446

.619

.030

.136

.408

.297

.273

.238

.096

.509

.408

.448

.124

.036

.792

.690

0.637

0.087

0.877

0.217

0.062

0.127

0.176

0.480

0.250

0.098

0.129

0.108

0.217

0.527

0.106

0.046

0.153

0.156

0.369

0.505

RES4

V25

8.817

0.147

0.186

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

DIR13
RESl
RESZ
RES3
RES4

RESS

RES6

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V21
V22
V23
V24
V25

V26

V27

-0.006
0.161
-0.460
0.529
~0.133
0.187
0.023
0.155
-0.367
0.005
0.161
0.076
-0.180
0.394
0.577
0.470
-0.076
0.595
0.287

-0.292

DIR13
V21
8.558
0.290
0.608
0.007
0.005
-0.191

-0.220

-0.250
-0.090
-0.254
0.578
-0.172
0.106
0.058
-0.S49
-0.081
0.217
-0.423
0.655
0.150
0.213
0.267
-0.928
-0.270
0.198
-0.879

-0.791

RESl
V22
8.102
0.885
0.035
0.001

—0.079

—0.235

~188-

-0.096

0.024

0.280

0.323

-0.272

0.129

-0.195

0.444

0.838

-0.486

0.215

0.638

-1.186

-0.264

—0.215

-0.176

-0.694

-0.129

0.121

0.340

RES2

V23

8.396
0.301
-0.070
0.069

0.448

-0.118

-0.655

0.598

-0.495

-0.527

-0.206

0.471

0.114

0.232

0.693

0.204

0.188

0.119

-0.017

-0.129

-0.244

-0.424

-0.640

-0.190

-0.611

RES3

V24

8.289
-0.449

-0.168

0.427

RES7
.322
RESB
.507
RES9
.269
RESlO
.208
RESll
.480
WTE1
.252
WTE2
.468
WTE3
.659
WTE4
.268
WTES
.252
WTE6
.021
WTE7
.481
WTE8
.283
WTE9
.169
WTE10
.722
WTE11
.121
VARl
.744
VAR2
.065
VAR3
.077
VAR4
.191
VARS
.127
VAR6
.247
VAR7
.146
BUR1
.065
BUR2
.448
BUR3
.656
BUR4
.263
BURS
.270

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

.237

.533

.004

.265

.448

.260

.429

.300

.008

.937

.075

.534

.221

.096

.274

.111

.011

.324

.060

.278

.327

.007

.370

.353

.663

.199

.135

.004

-189-

.691

.342

.190

.595

.172

.744

.401

.403

.099

.589

.632

.280

.346

.526

.442

.409

.019

.747

.361

.348

.394

.070

.282

.334

.149

.108

.088

.174

.515
.828
.335
.282
.281
.352
.366
.001
.223
.468
.669
.555
.410
.153
.370
.550
.109
.655
.247
.230
.456
.007
.245
.893
.333
.016
.847

.073

.282

.195

.229

.474

.132

.149

.108

.943

.144

.526

.181

.630

.333

.341

.093

.920

.316

.549

.342

.057

.009

.457

.387

.012

.022

.118

.681

.097

0.038

0.155

0.172

0.841

0.366

0.032

0.738

0.681

0.047

0.159

0.073

0.146

0.374

0.073

0.246

0.261

0.231

RES9

V30

8.506

0.372

0.306

0.104

0.259

0.260

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

RESS
RES6
RES7
RES8
RESQ
RESlO
RESll
WTE1
WTE2

WTE3

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V26
V27
V28
V29
V30
V31
V32
V33
V34

V35

-0.026

-0.470

0.558

0.314

1.002

0.122

-0.441

0.633

-0.415

-0.384

0.303

-0.204

-0.525

-0.030

-0.037

-0.554

0.481

RESS

V26
8.254
-0.131
0.472
-0.014
0.390
-0.234
-0.326
-0.027
0.393

-0.145

0.130

-0.981

-0.015

-0.115

0.411

-0.628

-0.141

-0.450

-0.236

-0.453

-0.048

0.110

-0.099

-0.297

0.371

0.328

-0.266

RES6

V27

8.677

-0.198

-0.403

0.210

0.035

-0.268

0.120

-0.340

0.084

-l90-

0.027

-0.452

0.245

-0.598

0.073

-0.047

-0.382

0.222

-0.651

-0.135

-0.002

-0.324

0.052

-0.622

0.772

0.661

-0.082

RES7

V28

8.099
-0.733
0.263
0.481
0.179
-0.467

-0.641

-0.154

-0.571

-0.630

-0.218

-0.182

0.324

—0.391

-0.116

0.282

0.168

0.222

-0.159

0.393

-0.S73

-0.746

0.393

—0.064

0.650

RESB

V29

8.321
-0.553

-0.642

-0.041

0.171

0.090

0.940

WTE4
0.093
WTES
0.477
WTE6
0.296
WTE7
0.206
WTE8
0.043
WTE9
0.019
WTE10
0.059
WTE11
0.484
VARl
0.179
VAR2
0.123
VAR3
0.231
VAR4
0.206
VARS
0.465
VAR6
0.0783
VAR7
0.33C)
BUR1
0.3222
BUR2
0.5423
BUR3
0.3656
BUR4
0.0658
BURS
0.3655
BUR6
0.7236
BUR7
0.2(14
BUR8
0.416)
BUR9
0.546
BUR10
0.494
BUR11
0.162
BUR12
L162
BUR13
0261

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

.209

.271

.543

.705

.213

.025

.478

.008

.018

.075

.128

.660

.000

.442

.043

.081

.142

.314

.610

.122

.148

.307

.364

.323

.681

.303

.184

.662

~191-

.180

.210

.286

.023

.294

.646

.776

.413

.091

.111

.087

.067

.253

.613

.338

.608

.359

.104

.161

.209

.439

.595

.451

.039

.376

.317

.255

.443

.311

.085

.430

.266

.042

.252

.424

.068

.247

.708

.139

.355

.049

.160

.519

.306

.162

.160

.165

.458

.494

.359

.416

.166

.282

.147

.231

.074

.257

.204

.654

.238

.183

.023

.009

.257

.163

.082

.416

.416

.043

.804

.617

.139

.446

.404

.356

.083

.050

.165

.257

.006

.308

.076

.763

.031

BUR14
0.089
BUR15
0.324
BUR16
0.183
BUR17
0.167
BUR18
0.167
BUR19
0.237
BUR20
0.079
WFR1
0.046
WFR2
0.396
WTE3
V35
RESlO
RESll
WTE1
WTE2
WTE3
9.345
WTE4
0.230
WTES
0.059
WTE6
0.132
WTE7
0.317
WTE8
0.224
WTE9
0.524
WTE10
0.276
WTE11
0.330
VAR1
0.320
VAR2
0.452
VAR3
0.353
VAR4
0.325
VARS
0.091

V64

V65

V66

V67

V68

V69

V70

V71

V72

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

RES

OOOl—‘CD

O

.220

.079

.116

.046

.541

.636

.508

.502

.213

10

V31

.070

.075

.064

.086

.038

.435

.844

.564

.377

.540

.057

.154

.041

.110

.537

.163

.236

.147

RES

0000)

O

.382

.173

.176

.125

.341

.071

.104

.193

.092

11

V32

.174

.540

.521

.215

.525

.009

.385

.142

.179

.220

.646

.209

.003

.121

.166

.126

.085

-l92-

.057

.356

.157

.141

.945

.267

.049

.494

.082

WTE1

V33

.917

.391

.903

.161

.186

.116

.025

.795

.429

.396

.058

.217

.302

.487

.534

.399

.310

.184

.304

.338

.446

.120

.214

.484

.057

WTE2

V34

.090
.314

.068

.084

.016

.762

.440

.069

.345

.214

.227

.211

.159

.082

.193

1.079

0.290

0.107

0.356

0.624

0.197

0.234

0.313

0.458

0.507

0.010

0.486

0.043

0.243

0.257

0.004

0.363

0.221

0.030

0.432

0.155

1.086

0.269

0.135

WTE8

V40

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

WTE4
WTES
WTE6

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V36
V37
V38

0.015

0.108

-0.235

0.090

-0.530

-0.654

0.608

-0.303

-0.631

0.665

-0.214

-0.114

0.131

-0.208

0.446

0.209

-0.333

0.654

-0.031

-0.068

-0.252

0.678

-0.222

0.246

WTE4
V36
8.049

0.567
0.237

-0.406

-0.093

0.400

0.314

0.449

-1.055

-0.471

-0.438

0.404

-0.096

-0.031

-0.040

0.598

0.216

0.061

-0.130

-0.143

0.187

0.330

0.384

-0.800

0.318

-0.252

-0.088

WTES

V37

8.417
0.389

-193-

0.087

-0.137

-0.274

-0.450

—0.161

-0.111

-0.825

0.301

0.382

0.183

0.456

-0.131

0.297

0.399

-0.S61

-0.017

-0.290

-0.778

-0.389

-0.116

0.299

0.056

0.069

-0.057

WTE6

V38

8.249

-0.281

-0.447

-0.018

0.891

-0.400

0.430

-0.199

-0.505

0.234

-0.030

-0.173

0.022

-0.664

-0.172

0.450

0.236

0.181

0.277

-0.373

0.673

-0.080

0.636

-0.135

0.009

WTE7

V39

.801

.069

.090

.583

.339

.159

.697

.450

.051

.785

.472

.276

.260

.723

.463

.732

.650

.180

.237

.662

.054

.527

.920

.250

.605

.845

.165

.336

WTE7
WTE8

WTE9

WTE10

WTE11

VAR1

VAR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

V39
V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

.148
.045

.062

.284

.218

.105

.150

.468

.552

.161

.346

.758

.043

.641

.140

.013

.296

.609

.444

.310

.053

.587

.351

.449

.153

.004

.497

.089

.800

-l94-

.424
.586

.696

.491

.302

.198

.360

.047

.986

.049

.379

.646

.429

.087

.064

.313

.093

.024

.016

.003

.553

.653

.188

.372

.456

.867

.151

.015

.893

.012
.141

.090

.290

.263

.332

.238

.306

.171

.409

.818

.097

.121

.137

.661

.093

.266

.158

.083

.581

.244

.203

.488

.115

.267

.329

.022

.046

.304

.507
.165

.743

.259

.654

.130

.720

.895

.355

.753

.129

.194

.028

.106

.604

.120

.611

.108

.341

.396

.241

.049

.051

.138

.366

.447

.238

.757

.406

BUR18
0.004
BUR19
0.100
BUR20
0.620
WFR1
0.199
WFR2
0.037
VAR2
V45
WTE9
WTE10
WTE11
VAR1
VAR2
8.415
VAR3
0.272
VAR4
0.637
VARS
0.114
VAR6
0.077
VAR7
0.440
BUR1
0.022
BUR2
0.383
BUR3
0.068
BUR4
0.531
BURS
0.213
BUR6
0.272
BUR7
0.471
BUR8
0.202
BUR9
0.106
BUR10
0.260
BUR11
0.452
BUR12
0.312

V68

V69

V70

V71

V72

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

.369

.264

.626

.193

.007

WTE9

OOOOW

V41

.325

.594

.310

.711

.064

.127

.160

.195

.860

.646

.225

.673

.275

.471

.019

.117

.541

.351

.036

.238

.359

.241

WTE1

.170

.081

.798

.039

.135

0

V42

-l95-

.118

.843

.113

.527

.191

.096

.842

.312

.570

.560

.061

.034

.487

.716

.739

.292

.536

.711

.038

.033

.111

WTE

-0.
—O.

.065

.084

.704

.272

.699

11

V43

.250

567

214

.321

.281

.520

.218

.185

.007

.406

.011

.355

.435

.266

.461

.091

.349

.011

.037

.875

.743

.634

.215

.042

.112

VAR1

V44

.446
.808

.160

.174

.807

.227

.148

.674

.289

.830

.717

.566

.307

.148

.350

.353

.097

.527

.608

0.473

0.293

0.145

0.048

0.370

0.105

0.872

0.155

0.205

0.114

VAR7

V50

8.070

0.569

0.757

0.723

0.181

0.388

0.509

0.618

0.444

0.054

0.140

1.086

0.371

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

-0.086
-0.373
0.482
-0.238
-0.199
-0.220
0.257
-0.294
0.339

0.201

VAR3
V46
8.380
0.347
0.812
0.679
0.301
-0.029
0.517
0.485
-0.038
0.052
-0.022
-0.S78
-0.227
*0.683
-0.179

—0.523

-0.920

-0.107

1.287

0.154

0.031

0.089

-0.179

-0.320

-0.859

-0.323

0.118

VAR4

V47

.666
.237
.288
.052

0000)

0.197

-0.220

0.621

0.345

-0.383

-0.745

-0.779

-0.394

-0.111

0.212

-0.612

0.235

~196-

0.276

-0.513

0.499

-0.046

0.022

-0.667

-0.S84

0.436

-0.476

-0.208

VARS

V48

8.984

-0.003

0.229

-0.041

0.445

-0.243

-0.589

-0.024

0.333

-0.302

-0.287

0.722

-0.271

0.470

-0.026

-0.738

-0.766

-0.061

0.074

-0.191

—0.324

0.500

-0.363

0.338

-0.085

VAR6

V49

8.501
-0.176

-0.087

-0.244

0.272

-0.889

-0.186

0.209

0.456

0.200

-0.139

-0.586

-0.769

0.350

BUR13
0.241

BUR14
0.312

BUR15
0.537

BUR16
0.764

BUR17
0.334

BUR18
0.564

BUR19
0.221

BUR20
0.242

WFR1
0.528
WFR2

0.022

BURS

VSS
BUR1
BUR2
BUR3
BUR4
BURS
7.767
BUR6
0.048
BUR7
0.236
BUR8
0.807
BUR9
0.253
BUR10
0.169
BUR11
0.084
BUR12
0.358
BUR13
0.481
BUR14
0.006
BUR15
0.450
BUR16
0.316
BUR17
0.335

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

0.342
0.269
-0.513
-0.574
0.204
-0.353
0.284
0.601
0.515

0.302

BUR1
V51
8.178
0.454
0.505
0.089
0.023
-0.860
-0.461
0.639
-0.047
0.222
0.138
0.734
-0.037
0.441
0.335

0.261

0.431

-0.151

0.010

-0.306

0.686

0.648

0.243

-0.025

0.105

0.257

0.062

BUR2

V52

8.288

-0.087

-0.049

0.747

-0.564

-0.792

0.074

-0.458

-0.013

0.100

0.156

0.019

-0.660

0.140

-0.042

-0.258

-l97-

—0.S41

-0.224

-1.079

-0.225

-0.380

0.157

-0.057

0.197

-0.187

0.083

BUR3

V53

8.257
.280
0.527

O

0.196

-0.335

-0.507

0.268

-0.430

0.246

-0.068

-0.130

0.694

0.039

-0.080

-0.235

-0.531

-0.225

-0.679

-0.497

0.108

-0.290

0.141

0.001

-0.331

0.280

BUR4

V54

8.466
0.251

0.700

—0.289

-0.201

0.107

-0.557

-0.267

-0.514

0.002

1.061

0.013

-0.144

0.346

 

0.468

0.128

0.706

0.490

0.098

BUR10

V60

8.191

0.211

0.600

0.195

0.648

0.634

0.143

0.059

0.483

0.183

0.178

0.323

0.292

BUR15

V65

BUR18

BUR19

BUR20

WFR1

WFR2

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

BUR11
BUR12
BUR13
BUR14

V68

V69

V70

V71

V72

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V61
V62
V63
V64

-0.441
0.131
-0.176
0.078

-0.132

BUR6
V56
7.921
0.469
0.347
0.194
0.139
0.216
-0.453
-0.504
0.480
0.575
-0.449
-0.516
0.128
-0.436
0.487

0.117

0.159

BUR11
V61

8.245
0.735
0.379
-0.023

-0.078

0.230

0.072

-0.045

-0.143

BUR7

V57

8.611

-0.098

-0.100

-0.434

0.470

0.533

-0.002

0.123

-0.634

-0.S90

-0.453

-0.359

0.189

-0.424

-0.024

0.452

BUR12

V62

8.118
-0.400
-0.129

-198-

-0.016

-0.341

0.656

0.421

-0.974

BUR8

V58

8.093

-0.237

0.463

-0.904

-0.126

0.290

0.381

-0.246

0.339

-0.041

-0.S92

-0.177

0.481

0.457

0.305

BUR13

V63

7.838
-0.466

0.246

0.390

-0.472

0.008

0.344

BUR9

V59

7.995
-0.144

0.156

0.820

-0.557

0.266

-0.250

-0.477

0.775

-0.350

-0.353

0.248

0.161

0.054

BUR14

V64

8.260

BUR15

.660

BUR16

.420

BUR17

.526

BUR18

.086

BUR19

.048

BUR20

.069

WFR1

.660

WFR2

.169

BUR20

V7

8.

0.

0.

10

20

31

42

0
BUR16
BUR17
BUR18
BUR19
BUR20
090
WFR1
258
WFR2
346

WFR1
WFR2

V65

V66

V67

V68

V69

V70

V71

V72

V66
V67
V68
V69
V70

V71

V72

V71
V72

0.380
0.261
-0.125
0.149
0.183
—0.008
—0.302

0.358

BUR16
V66

7.788
0.424
0.507
0.377
0.341

0.050

0.228

WFR1

V71
8.263
-0.237

BUR1

.584

.180

.061

.321

.459

.299

.708

.393

7

V67

.295
.294
.194
.145

.042

.267

WFR2
V72

8.

180

BENTLER-WEEKS STRUCTURAL REPRESENTATION:

NUMBER OF DEPENDENT VARIABLES
DEPENDENT V'S
DEPENDENT F'S

8
1

21
2

= 1

NUMBER OF INDEPENDENT VARIABLES =
INDEPENDENT V'S

11

22

33

44

INDEPENDENT V'S

INDEPENDENT V'S

INDEPENDENT V'S

1

12

23

34

-l99-

2

13

24

35

S
32
3

73
3

14

25

36

.251

.080

.747

.384

.478

.390

.888

.320

BUR18

V68

.195
.155
.215
.739

.797

15

26

37

16

27

38

0.313

0.369

0.027

0.079

0.515

0.207

0.571

0.173

BUR19

V69

8.017
0.142

0.379

0.175

17

28

39

18

29

40

19

30

41

 

INDEPENDENT V'S : 45 46 47 49 50 51 52

54 55

INDEPENDENT V'S : 56 57 58 59 60 61 62
64 6S

INDEPENDENT V'S : 66 67 68 69

INDEPENDENT E'S ° 8 21 32 43 48 70 71

INDEPENDENT D'S : 7

NUMBER OF FREE PARAMETERS = 144
NUMBER OF FIXED NONZERO PARAMETERS = 16

*** WARNING MESSAGES ABOVE, IF ANY, REFER TO THE MODEL PROVIDED.
CALCULATIONS FOR INDEPENDENCE MODEL NOW BEGIN.

*** WARNING MESSAGES ABOVE, IF ANY, REFER TO INDEPENDENCE MODEL.
CALCULATIONS FOR USER'S MODEL NOW BEGIN.

3RD STAGE OF COMPUTATION REQUIRED 957288 WORDS OF MEMORY.

PROGRAM ALLOCATED 40000000 WORDS

DETERMINANT OF INPUT MATRIX IS 0.20969D+63

PARAMETER CONDITION CODE

D7,D7 LINEARLY DEPENDENT ON OTHER PARAMETERS

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)

RESIDUAL COVARIANCE MATRIX (S-SIGMA)

PRE1 PRE2 PRE3 PRE4
PRES
V1 V2 V3 V4

V5

PRE1 V1 0.000

PRE2 V2 -0.372 0.000

PRE3 V3 0.079 -0.620 0.000

PRE4 V4 -0.363 0.126 -0.131 0.000

PRES V5 1.227 -0.592 1.014 0.289
0.000

PRE6 V6 0.050 -0.716 -0.708 0.204
0.763

PRE7 V7 -0.061 0.800 0.773 -0.291
0.366

PRE8 V8 0.221 —0.101 -0.653 -0.475
0.757

DIR1 V9 0.016 -0.l39 -0.091 0.260
0.353

DIR2 V10 -0.380 -0.506 -0.716 0.501
0.765

DIR3 V11 -0.226 0.220 -0.384 -0.040
0.657

-200-

S3

63

72

DIR4
.153
DIRS
.131
DIR6
.326
DIR7
.254
DIR8
.355
DIR9
.491
DIR10
.754
DIR11
.212
DIR12
.759
DIR13
.293
RESI
.336
REsz
.518
RES3
.289
RES4
.202
RESS
.280
RESS
.161
RES7
.333
REsa
.261
RES9
.037
RESIO
.838
RESll
.240
WTE1
.056
WTE2
.452
WTE3
.089
WTE4
.264 '
WTES
.414
WTE6
.379
WTE7
.241

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

. 560

.124

.566

.286

.225

.204

.117

.067

.530

.011

.337

.909

.670

.219

.093

.652

.163

.103

.729

. 032

.510

.639

.216

. 001

. 100

. 175

.204

.304

-201-

.533

.263

.345

.103

.336

.464

.194

.521

.265

.467

.216

.035

.415

.197

.112

.044

.597

.038

.386

.356

.480

.034

.602

. 509

.047

.145

.134

.148

.025

.236

.367

.223

.025

.091

.216

.949

.623

.712

.242

.393

.304

.070

.110

.710

.024

.194

. 753

.007

.007

.220

.057

.561

.667

.036

.758

.190

.444

.724

. 001

.323

.103

.456

.565

.198

. 089

.176

.148

.311

.437

.090

.242

.218

.032

.439

.221

.043

.669

.253

.708

.430

.540

.191

.110

.644

WTE8
.162
WTE9
.233
WTE10
.408
WTE11
.429
VAR1
.530
VAR2
.829
VAR3
.328
VAR4
.326
VARS
.290
VAR6
.456
VAR7
.476
BUR1
.270
BUR2
.844
BUR3
.514
BUR4
.675
BURS
.425
BUR6
.325
BUR7
.103
BUR8
.234
BUR9
.383
BUR10
.154
BUR11
.349
BUR12
.128
BUR13
.970
BUR14
.757
BUR15
.674
BUR16
.314
BUR17
.533

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

.302

.141

.744

.599

.273

.097

.106

.278

.101

.044

.186

.708

.611

.758

.153

.254

.040

.046

.475

.001

.975

.420

.648

.198

.103

.263

.092

.237

-202-

.357

.035

.354

.371

.247

.890

.502

.322

.155

.140

.254

.168

.176

.056

.522

.185

.301

.477

.682

.483

.095

.326

.146

.113

.487

.633

.102

.028

.333

.359

.049

.039

.043

.141

.155

.036

.227

.307

.509

.686

.000

.657

.343

.268

.163

.570

.981

.052

.060

.334

.893

.104

.305

.252

.603

.210

.948

.206

.267

.814

.040

.666

.335

.475

.491

.307

.201

.240

.127

.047

.601

.346

.512

.220

.281

.126

.316

.402

.757

.003

.665

.073

.412

.085

BUR18
0.446
BUR19
0.177
BUR20
0.020
WFR1
0.228
WFR2
0.812
DIR2
V10
PRE6
PRE7
PRE8
DIR1
DIR2
0.000
DIR3
0.016
DIR4
1.153
DIRS
0.264
DIR6
0.599
DIR7
0.394
DIR8
0.535
DIR9
0.201
DIR10
0.153
DIR11
0.027
DIR12
0.769
DIR13
0.703
RESl
0.196
RESZ
0.008
RES3
0.223
RES4
0.265
RESS
0.682
RES6
0.769

V68

V69

V70

V71

V72

V10

V11

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

.126

.979

.549

.260

.523

PRE6

V6

.000

.094

.682

.397

.300

.465

.696

.161

.276

.147

.279

.252

.585

.118

.041

.130

.351

.684

.059

.294

.356

.417

.141

.094

.243

.146

.238

PRE7

V7

-203-

.000

.181

.179

.173

.141

.830

.819

.242

.411

.466

.210

.170

.502

.623

.353

.579

.329

.081

.716

.199

.329

.812

.727

.040

.260

.638

PRE8

V8

.000

.311

.508

.026

.445

.338

.315

.187

.922

.051

.617

.081

.094

.078

.306

.413

.231

.083

.461

.522

.516

.119

.390

.284

.545

DIR1

V9

.000
.106

.320

.166

.373

.203

.120

.022

.528

.579

.120

.099

.510

.536

.778

.000

.203

.140

.082

 

RES7
.127
RESB
.259
RES9
.092
RESlO
.167
RESll
.313
WTE1
.071
WTE2
.235
WTE3
.598
WTE4
.461
WTES
.668
WTE6
.130
WTE7
.265
WTE8
.044
WTE9
.560
WTE10
.087
WTE11
.177
VAR1
.202
VAR2
.357
VAR3
.568
VAR4
.772
VARS
.270
VAR6
.615
VAR7
.434
BUR1
.009
BUR2
.092
BUR3
.639
BUR4
.085
BURS
.108

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

.037

.355

.305

.398

.004

.608

.515

.081

.537

.260

.440

.164

.018

.115

.240

.368

.508

.029

.046

.005

.146

.497

.173

.076

.421

.143

.126

.250

~204-

.357

.855

.259

.421

.177

.000

.543

.478

.695

.470

.177

.024

.690

.709

.129

.017

.975

.178

.602

.427

.470

.259

.072

.266

.342

.018

.534

.410

.470

.561

.764

.511

.403

.276

.369

.152

.282

.298

.663

.189

.008

.345

.716

.236

.053

.064

.030

.468

.413

.323

.261

.697

.162

.612

.110

.223

.149

.324

.057

.368

.067

.259

.236

.130

.265

.075

.361

.619

.064

.040

.070

.114

.134

.187

.284

.722

.413

.099

.059

.729

.822

.134

.348

.073

 

0.470

0.313

0.149

0.125

0.267

0.635

0.161

0.393

0.107

0.381

0.002

0.189

0.315

0.862

0.051

0.243

0.602

DIR7

V15

0.000

0.914

0.476

0.360

0.449

0.259

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

DIR3
DIR4
DIRS
DIR6
DIR7

DIR8

DIR9

DIR10

DIR11

DIR12

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V11
V12
V13
V14
V15
V16
V17
V18
V19

V20

0.645
-0.400
0.017
-0.434
0.336
-0.240
0.121
-0.263
0.163
0.229
—0.621
-0.352
-0.634
-0.232
-0.154
-0.472

0.399

DIR3
V11
0.000
0.187
0.207
0.247
0.177
-0.028
-0.693
-0.990

0.162

-0.073

0.287

-0.533

0.352

0.432

0.423

-0.555

0.016

0.194

0.903

0.496

-0.748

0.561

0.163

0.117

-0.460

0.213

0.192

DIR4

V12

0.000

0.168

—0.169

0.248

0.221

0.293

-0.260

-0.257

-0.138

0.084

0.149

0.632

0.234

0.178

-0.098

-0.487

-0.031

-0.092

-0.212

0.586

0.542

-0.709

0.249

0.158

-0.284

4.758

DIRS

V13

0.000
-0.428
-0.098
-0.932
-0.572
—0.081

-0.291

-0.162

-0.253

0.281

0.351

-0.217

-0.014

-0.025

-0.314

0.159

0.245

0.598

0.412

0.054

0.048

0.186

0.491

-0.172

0.152

DIR6

V14

0.000
-0.035

-0.420

-0.321

-0.213

-0.610

0.264

DIR13
.024
RESl
.379
RES2
.126
RES3
.300
RES4
.095
RESS
.157
RES6
.005
RES7
.287
RESB
.079
RES9
.161
RESlO
.092
RESll
.577
WTE1
.393
WTE2
.183
WTE3
.565
WTE4
.595
WTES
.083
WTE6
.432
WTE7
.466
WTE8
.217
WTE9
.529
WTE10
.044
WTE11
.101
VAR1
.377
VAR2
.011
VAR3
.455
VAR4
.078
VARS
.465

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

.140
.439
.257
.071
.801
.004
.331
.252
.217
.455
.403
.967
.689
.580
.556
.436
.300
.243
.968
.162
.323
.264
.298
.039
.465
.341
.314

.141

-206-

.431

.021

.374

.311

.554

.104

.273

.425

.321

.111

.082

.681

.208

.052

.324

.023

.303

.173

.888

.177

.585

.061

.180

.550

.018

.230

.362

.281

.227

.156

.593

.394

.027

.007

.218

.145

.182

.404

.006

.461

.317

.208

.198

.308

.160

.771

.319

.206

.094

.334

.243

.638

.553

.414

.044

.338

.460

.151

.531

.386

.433

.402

.016

.693

.342

.405

.261

.522

.369

.059

.056

.221

.055

.491

.029

.206

.247

.417

.528

.053

.045

.232

.130

.689

.415

.199

.601

.121

.061

.577

.073

.145

.163

.507

.079

.145

.098

.385

.575

.806

.025

.368

.060

.580

.308

.294

.083

.364

DIR12

V20

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

DIR8
DIR9
DIR10

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V16
V17
V18

0.041

0.261

0.106

0.368

0.647

-0.567

-0.676

0.486

-0.049

-0.105

0.278

0.357

0.125

0.276

0.039

-0.007

-0.574

-0.324

-0.713

-0.286

0.115

0.129

0.157

0.481

DIR8

V16

0.000

0.214
-0.674

-0.425

0.357

-0.S38

0.761

-0.291

0.517

0.409

0.830

0.015

0.284

0.605

-0.163

0.342

-0.212

0.538

-0.034

0.746

0.087

0.323

-0.229

0.489

-0.689

0.496

-0.040

DIR9

V17

0.000
0.349

-207-

—0.963

-0.026

0.170

—0.145

0.297

0.055

0.026

-0.061

0.284

0.283

-0.204

0.428

0.347

0.304

0.240

0.141

-0.267

0.318

-0.239

0.738

0.064

0.683

1.364

-0.710

DIR10

V18

0.000

0.819

0.025

-0.299

—0.381

-0.779

-0.219

-0.012

-0.109

0.239

0.596

—0.933

0.186

-0.135

-0.141

0.006

-0.616

-0.171

—0.256

-0.265

-0.423

0.193

-0.298

-0.129

3.722

DIR11

V19

 

DIR11

0.000
0.249
0.381
0.084
0H721.
0.9636
0.0233
0.22L7
0.1417
0.2554
0.4221
(3.0597
0.:LJ.7
0.3'76
0.5416
0.8C)2
0.0373
0.2C)7
0.3C)2
0.93¢4
0.3223
0.9226
0.7222
0.1653
0.1C)3
0.13'7
0.14¢1

0.473

DIR12

DIR13

RESl

RESZ

RES3

RES4

RESS

RES6

RES7

RESB

RES9

RESlO

RESll

WTE1

WTE2

WTE3

WTE4

WTES

WTE6

WTE7

WTE8

WTE9

WTE10

WTE11

VAR1

VAR2

VAR3

VAR4

V19
V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

-0.
-0.

553
768

.179

.179

.033

.756

.356

.088

.078

.059

.475

.389

.291

.116

.389

.449

.122

.190

.435

.471

.250

.025

.159

.714

.306

.307

.207

.265

.017

-0.669
0.435

-0.251

-0.281

0.093

0.157

0.083

0.007

0.085

0.616

-0.163

-0.S71

0.324

0.146

0.205

0.008

-0.611

-0.103

-0.740

-0.225

0.486

0.603

-0.292

0.099

0.301

0.313

0.454

-0.723

-0.139

-208-

.637
.271

.740

.412

.055

.187

.273

.013

.292

.048

.097

.192

.344

.344

.077

.026

.531

.280

.103

.174

.308

.172

.886

.061

.115

.508

.133

.090

.266

.000
.051

.803

.148

.109

.307

.011

.159

.428

.788

.028

.475

.287

.108

.005

.265

.401

.046

.446

.619

.030

.136

.408

.297

.273

.238

.096

.509

.408

 

VARS
0.067
VAR6
0.395
VAR7
0.039
BUR1
0.016
BUR2
0.317
BUR3
0.637
BUR4
0.087
BURS
0.877
BUR6
0.217
BUR7
0.062
BUR8
0.127
BUR9
0.176
BUR10
0.480
BUR11
0.250
BUR12
0.098
BUR13
0.129
BUR14
0.108
BUR15
0.217
BUR16
0.527
BUR17
0.106
BUR18
0.046
BUR19
0.153
BUR20
0.156
WFR1
0.301
WFR2
0.507
RES4
V25
DIR13

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V21

-0.S96

0.125

0.043

-0.322

0.540

-0.006

0.161

-0.460

0.529

-0.133

0.187

0.023

0.155

-0.367

0.005

0.161

0.076

-0.180

0.394

0.577

0.470

-0.076

0.595

0.118

-0.297

DIR13

V21

-0.027

-0.440

0.644

0.066

0.008

0.296

-0.250

-0.090

-0.254

0.578

-0.172

0.106

0.058

-0.S49

-0.081

0.217

-0.423

0.655

0.150

0.213

0.267

-0.928

-0.270

0.198

-0.552

-0.781

RESl

V22

0.412

0.126

0.161

0.124

0.405

-0.096

0.024

0.280

0.323

-0.272

0.129

-0.195

0.444

0.838

-0.486

0.215

0.638

-1.186

-0.264

-0.215

-0.176

-0.694

-0.129

0.084

0.339

RES2

V23

0.448

-0.124

-0.036

-0.792

-0.690

—0.118

-0.655

0.598

-0.495

-0.527

-0.206

0.471

0.114

0.232

0.693

0.204

0.188

0.119

-0.017

-0.129

-0.244

-0.424

-0.640

-0.305

-0.614

RES3

V24

 

A"

.000

.147

.186

.322

.507

.269

.208

.024

.252

.468

.659

.268

.252

.021

.481

.283

.169

.722

RESl
RES2
RES3
RES4
RESS
RES6
RES7

RESB

RES9

RESlO

RESll

WTE1

WTE2

WTE3

WTE4

WTES

WTE6

WTE7

WTE8

WTE9

WTE10

WTE11

.121

.744

.065

.077

.191

.127

.247

.146

.065

VAR1

VAR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

.290

.608

.007

.005

.191

.220

.237

.533

.004

.265

.448

.260

.429

.300

.008

.937

.075

.534

.221

.096

.274

.111

.011

.324

.060

.278

.327

.007

.370

.353

0000

-210-

.000

.885

.035

.001

.079

.235

.691

.342

.190

.595

.128

.744

.401

.403

.099

.589

.632

.280

.346

.526

.442

.409

.019

.747

.361

.348

.394

.070

.282

.334

.000

.301

.070

.069

.448

.515

.828

.335

.282

.239

.352

.366

.001

.223

.468

.669

.555

.410

.153

.370

.550

.109

.655

.247

.230

.456

.007

.245

.893

.000
.449

.168

.427

.282

.195

.229

.474

.099

.149

.108

.943

.144

.526

.181

.630

.333

.341

.093

.920

.316

.549

.342

.057

.009

.457

.387

.012

 

0.448

0.656

0.263

0.270

0.038

0.155

0.172

0.841

0.366

0.032

0.738

0.681

0.047

0.159

0.073

0.146

0.374

0.073

0.246

0.076

0.237

RES9

V30

0.000

0.372

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

RESS
RES6
RES7
RESB
RES9

RESlO

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V26
V27
V28
V29
V30

V31

-0.663

0.199

-0.135

-0.004

-0.026

-0.470

0.558

0.314

1.002

0.122

-0.441

0.633

-0.415

-0.384

0.303

-0.204

—0.S2S

-0.030

-0.037

-0.S63

0.481

RESS

V26

0.000
-0.131
0.472
-0.014
0.390

-0.234

0.149

0.108

0.088

-0.174

0.130

-0.981

-0.015

-0.115

0.411

-0.628

-0.141

-0.450

-0.236

-0.453

-0.048

0.110

-0.099

-0.297

0.371

0.344

-0.266

RES6

V27
0.000
-0.198
-0.403
0.210

0.035

-2H-

0.333

0.016

-0.847

0.073

0.027

-0.452

0.245

-0.S98

0.073

-0.047

-0.382

0.222

-0.651

-0.135

-0.002

-0.324

0.052

-0.622

0.772

0.470

-0.087

RES7

V28

0.000
-0.733
0.263

0.481

-0.022

-0.118

—0.681

-0.097

-0.571

—0.630

-0.218

-0.182

0.324

-0.391

-0.116

0.282

0.168

0.222

-0.159

0.393

-0.S73

-0.746

0.393

-0.051

0.650

RESB

V29

0.000
-0.553

-0.642

 

RESll
.065
WTE1
.104
WTE2
.259
WTE3
.260
WTE4
.093
WTES
.477
WTE6
.296
WTE7
.206
WTE8
.043
WTE9
.019
WTE10
.059
WTE11
.484
VAR1
.179
VAR2
.123
VAR3
.231
VAR4
.206
VARS
.465
VAR6
.078
VAR7
.330
BUR1
.322
BUR2
.542
BUR3
.366
BUR4
.068
BURS
.365
BUR6
.736
BUR7
.204
BUR8
.410
BUR9
.546

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

.151

.027

.393

.145

.209

.271

.543

.705

.213

.025

.478

.008

.018

.075

.128

.660

.000

.442

.043

.081

.142

.314

.610

.122

.148

.307

.364

.323

-2l2-

.069

.120

.340

.084

.180

.210

.286

.023

.294

.646

.776

.413

.091

.111

.087

.067

.253

.613

.338

.608

.359

.104

.161

.209

.439

.595

.451

.039

.287

.467

.641

.154

.311

.085

.430

.266

.042

.252

.424

.068

.247

.708

.139

.355

.049

.160

.519

.306

.162

.160

.165

.458

.494

.359

.416

.166

.259

.171

.090

.940

.257

.204

.654

.238

.183

.023

.009

.257

.163

.082

.416

.416

.043

.804

.617

.139

.446

.404

.356

.083

.050

.165

.257

.006

 

 

BUR10
0.494
BUR11
0.162
BUR12
1.162
BUR13
0.261
BUR14
0.089
BUR15
0.324
BUR16
0.183
BUR17
0.167
BUR18
0.167
BUR19
0.237
BUR20
0.079
WFR1
0.091
WFR2
0.401

WTE3

V35
RESIO
RESll
WTE1
WTE2
WTE3
0.000
WTE4
0.230
WTES
0.059
WTE6
0.132
WTE7
0.317
WTE8
0.224
WTE9
0.524
WTE10
0.276
WTE11
0.043
VAR1
0.320

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V31
V32
V33
V34
V35
V36
V37
V38
V39
V40
V41
V42

V43

V44

0.681
-0.303
0.184
-0.662
0.220
-0.079
0.116
0.046
-0.541
0.636
0.508
-0.326

0.218

RESlO
V31

0.000
0.174
0.064
0.086
0.038
0.435
-0.844
-0.564
0.377
-0.S40
0.057
0.154

0.041

-0.110

0.376

-0.317

-0.255

0.443

0.382

-0.173

-0.176

0.125

0.341

-0.071

0.104

0.267

-0.090

RESll

V32

-0.012

0.540

0.521

-0.215

0.525

-0.009

-0.385

0.142

-0.179

-0.220

-0.646

-0.209

0.003

0.282

0.147

-0.231

-1.074

0.057

0.356

-0.157

0.141

-0.945

-0.267

0.049

-0.454

0.083

WTE1

V33

0.000

0.391

-0.903

0.161

0.186

-0.116

-0.025

0.795

-0.429

-0.396

0.126

-0.217

—0.308

-0.076

0.763

0.031

-0.310

-1.184

0.304

-0.338

0.446

-0.l20

0.214

0.404

-0.060

WTE2

V34

0.000
-0.314

-0.068

-0.084

-0.016

0.762

0.440

-0.069

-0.345

-0.055

0.227

 

 

 

VAR2 V45 0.537 0.121 -0.302 0.211

.452
VAR3 V46 0.163 -0.166 -0.487 -0.159
.353
VAR4 V47 -0.236 -0.126 0.534 -0.082 -
.325
VARS V48 -0.147 ~0.085 0.399 0.193
.091
VAR6 V49 0.015 -0.406 0.087 -0.281 —
.079
VAR7 V50 0.108 -0.093 -0.137 -0.447
.290
BUR1 V51 -0.235 0.400 -0.274 -0.018 -
.107
BUR2 V52 0.090 0.314 -0.450 0.891
.356
BUR3 V53 -0.530 0.449 -0.161 -0.400 -
.624
BUR4 V54 -0.654 -1.055 -0.111 0.430
.197
BURS VSS 0.608 -0.471 ~0.825 -0.199
.234
BUR6 V56 -0.303 -0.438 0.301 -0.505
.313
BUR7 V57 -0.631 0.404 0.382 0.234 -
.458
BUR8 V58 0.665 -0.096 0.183 -0.030
.507
BUR9 V59 -0.214 —0.031 0.456 -0.173
.010
BUR10 V60 -0.114 -0.040 -0.131 0.022
.486
BUR11 V61 0.131 0.598 0.297 ~0.664 -
.043
BUR12 V62 -0.208 0.216 0.399 -0.172 -
.243
BUR13 V63 0.446 0.061 -0.S61 0.450 -
.257
BUR14 V64 0.209 -0.130 -0.017 0.236
.004
BUR15 V65 —0.333 -0.143 -0.290 0.181
.363
BUR16 V66 0.654 0.187 -0.778 0.277
.221
BUR17 V67 -0.031 0.330 -0.389 -0.373 -
.030
BUR18 V68 -0.068 0.384 -0.116 0.673
.432
BUR19 V69 -0.252 -0.800 0.299 -0.080 -
.155
BUR20 V70 0.678 0.318 0.056 0.636
.086
WFR1 V71 -0.555 -0.332 0.084 -0.192
.330
WFR2 V72 0.236 -0.090 -0.056 0.007 -
.133

~2M~

 

 

 

 

 

 

 

WTE4 WTES WTE6 WTE7

WTE8
V36 V37 V38 V39
V40
WTE4 V36 0.000
WTES V37 0.567 0.000
WTE6 V38 0.237 0.389 0.000
WTE7 V39 0.148 -0.424 0.012 0.000
WTE8 V40 -0.045 0.586 -0.141 0.165
0.000
WTE9 V41 -0.062 0.696 -0.090 -0.743 -
0.069
WTE10 V42 -0.284 -0.491 -0.290 -0.259
0.090
WTE11 V43 0.141 0.208 -0.088 -0.034 -
0.044
VAR1 V44 -0.105 0.198 -0.332 -0.130
0.339
VAR2 V45 0.150 -0.360 0.238 0.720 -
0.159
VAR3 V46 -0.468 0.047 —0.306 -0.895 -
0.697
VAR4 V47 -0.552 0.986 -0.171 0.355 -
0.450
VARS V48 0.161 -0.049 -0.409 -0.753
0.051
VAR6 V49 -0.346 -0.379 -0.818 0.129
0.785
VAR7 V50 0.758 -0.646 0.097 0.194 -
0.472
BUR1 V51 0.043 -0.429 -0.121 -0.028 -
0.276
BUR2 V52 0.641 0.087 -0.137 0.106
0.260
BUR3 V53 -0.140 -0.064 -0.661 0.604
0.723
BUR4 V54 0.013 -0.313 0.093 0.120 -
0.463
BURS VSS 0.296 -0.093 -0.266 0.611 -
0.732
BUR6 V56 -0.609 -0.024 0.158 0.108
0.650
BUR7 V57 0.444 0.016 0.083 0.341
0.180
BUR8 V58 0.310 0.003 0.581 -0.396 -
0.237
BUR9 V59 0.053 -0.553 0.244 0.241
0.662
BUR10 V60 -0.587 -0.653 -0.203 -0.049
0.054
BUR11 V61 -0.351 0.188 0.488 0.051 -
0.527
BUR12 V62 -0.449 -0.372 -0.115 0.138
0.920
BUR13 V63 -0.153 0.456 -0.267 0.366 -
0.250

-215-

 

0.605

0.845

0.165

0.336

0.004

0.100

0.620

0.332

0.041

VAR2

V45

0.000

0.272

0.637

0.070

0.077

0.440

0.022

0.383

0.068

0.531

0.213

0.272

0.471

0.202

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

WTE9

WTE10

WTE11

VAR1

VAR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

V64

V65

V66

V67

V68

V69

V70

V71

V72

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

0.004

-0.497

-0.089

-0.800

-0.369

-0.264

-0.626

-0.209

0.007

WTE9

V41

0.000

0.594

-0.166

0.711

0.064

0.127

0.160

0.195

-0.860

0.646

0.225

0.673

0.275

0.471

-0.019

0.117

0.541

0.351

-0.867

-0.151

0.015

0.893

0.170

0.081

-0.798

0.070

-0.131

WTE10

V42

0.000

0.002

-0.113

0.527

-0.191

-0.096

-0.842

0.312

0.570

0.560

-0.061

-0.034

0.487

0.716

0.739

—0.292

-0.536

-2l6-

0.329

0.022

-0.046

-0.304

0.065

-0.084

0.704

0.197

0.697

WTE11

V43

-0.028

-0.567

-0.214

-0.321

—0.281

0.520

0.218

-0.185

-0.007

-0.406

-0.011

0.355

-0.435

-0.266

-0.461

0.091

-0.447

0.238

0.757

0.406

0.743

-0.634

0.215

0.174

-0.108

VAR1

V44

0.000
0.808

-0.160

0.174

0.016

0.227

-0.148

-0.674

0.289

-0.830

-0.717

-0.S66

-0.307

0.148

0.350

 

0.106

0.260

0.452

0.312

0.473

0.293

0.145

0.048

0.370

0.105

0.872

0.155

0.197

0.114

VAR7

V50

0.000

0.569

0.757

0.723

0.181

0.388

0.509

0.618

0.444

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

VAR3
VAR4
VARS
VAR6
VAR7
BUR1
BUR2
BUR3
BUR4
BURS
BUR6

BUR7

BUR8

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V46
V47
V48
V49
V50
V51
V52
V53
V54
V55
V56

V57

V58

-0.036
0.238
0.359
0.241

-0.086

-0.373
0.482

-0.238

-0.199

-0.220
0.257

-0.294
0.238

0.198

VAR3
V46
0.000
0.347
0.038
0.679
0.301
-0.029
0.517
0.485
-0.038
0.052
-0.022

-0.578

-0.227

-0.711

0.038

0.033

0.111

-0.107

1.287

0.154

0.031

0.089

-0.179

-0.320

-0.859

-0.143

0.123

VAR4

V47

0.000

-0.130

0.288

0.052

0.197

-0.220

0.621

0.345

-0.383

-0.745

-0.779

-0.394

-217-

0.349

0.011

-0.037

-0.875

0.276

-0.513

0.499

-0.046

0.022

-0.667

-0.584

0.436

-0.S35

-0.210

VARS

V48

0.001

0.034

0.130

-0.041

0.445

-0.243

-0.589

-0.024

0.333

-0.302

-0.287

0.353

0.097

-0.527

0.608

-0.738

-0.766

-0.061

0.074

-0.191

-0.324

0.500

-0.363

0.201

-0.089

VAR6

V49

0.000
-0.176

-0.087

-0.244

0.272

—0.889

-0.186

0.209

0.456

0.200

BUR9
0.054
BUR10
0.140
BUR11
1.086
BUR12
0.371
BUR13
0.241
BUR14
0.312
BUR15
0.537
BUR16
0.764
BUR17
0.334
BUR18
0.564
BUR19
0.221
BUR20
0.242
WFR1
0.545
WFR2
0.023

BURS

V55
BUR1
BUR2
BUR3
BUR4
BURS
0.000
BUR6
0.048
BUR7
0.236
BUR8
0.807
BUR9
0.253
BUR10
0.169
BUR11
0.084
BUR12
0.358
BUR13
0.481

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V51
V52
V53
V54
V55
V56
V57
V58
V59
V60
V61

V62

V63

-0.683
-0.179
-0.523
-0.920
0.342
0.269
-0.513
—0.S74
0.204
-0.353
0.284
0.601
0.380

0.297

BUR1
V51
0.000
0.454
0.505
0.089
0.023
-0.860
-0.461
0.639
-0.047
0.222
0.138

0.734

-0.037

-0.111

0.212

-0.612

0.235

-0.151

0.010

-0.306

0.686

0.648

0.243

-0.025

0.105

0.193

0.060

BUR2

V52

0.000

-0.087

-0.049

0.747

-0.564

-0.792

0.074

-0.458

-0.013

0.100

0.156

0.019

-218-

0.722

-0.271

0.470

-0.026

~0.S41

*0.224

-1.079

-0.225

-0.380

0.157

-0.057

0.197

-0.216

0.082

BUR3

V53

0.000
.280
0.527

O

0.196

-0.335

-0.507

0.268

-0.430

0.246

-0.068

-0.130

-0.139
-0.586
-0.769
0.350
-0.531
-0.225
-0.679
-0.497
0.108
-0.290
0.141
0.001
—0.324

0.281

BUR4

V54

0.000
0.251

0.700
-0.289
-0.201

0.107
-0.557
-0.267
-0.514

0.002

 

 

0.006

0.450

0.316

0.335

0.468

0.128

0.038

0.538

0.100

BUR10

V60

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

BUR6
BUR7
BUR8
BUR9

BUR10
.000
BUR11
.211
BUR12
.600
BUR13
.195
BUR14
.648
BUR15
.634
BUR16
.143
BUR17
.059
BUR18
.483
BUR19
.183
BUR20
.050

WFR1

.308

WFR2

.291

V64

V65

V66

V67

V68

V69

V70

V71

V72

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

0.441
0.335
0.261
0.431
-0.441
0.131
0.058
0.093

-0.131

BUR6
V56
0.000
0.469
0.347
0.194
0.139
0.216
-0.453
-0.504
0.480
0.575
-0.449
-0.516
0.128
-0.436
0.030

0.088

0.159

-0.660
0.140
-0.042
-0.258
-0.078
0.230
~0.095
fi-0.056

-0.144

BUR7

V57
0.000
-0.098
-0.100
-0.434
0.470
0.533
-0.002
0.123
-0.634
-0.590
-0.453
-0.359
0.189
0.012

0.004

0.453

-2|9-

0.694

0.039

-0.080

-0.235

—0.016

-0.341

-0.090

0.373

—0.975

BUR8

V58

0.000

-0.237

0.463

-0.904

-0.126

0.290

0.381

-0.246

0.339

-0.041

-0.S92

-0.177

0.090

0.431

0.304

1.061

0.013

-0.144

0.346

0.246

0.390

0.067

0.043

0.345

BUR9

V59

0.000
-0.144

0.156

0.820

-0.557

0.266

-0.250

-0.477

0.775

-0.350

-0.353

-0.002

0.145

0.053

BUR15

V65
BUR11
BUR12
BUR13
BUR14
BUR15
0.000
BUR16
0.420
BUR17
0.526
BUR18
0.086
BUR19
0.048
BUR20
0.014
WFR1
0.665
WFR2
0.169

BUR20

V70
BUR16
BUR17
BUR18
BUR19
BUR20
0.029
WFR1
0.236
WFR2
0.346

WFR1
WFR2

0.3211

0.3298

V61
V62
V63
V64
V65
V66
V67
V68
V69
V70

V71

V72

V66
V67
V68
V69
V70

V71

V72

V71
V72

BUR11

V61

.000
.735
.379
.023
.380

00000

O

.261

-0.125

0.149

0.183

0.010

-0.301

0.358

BUR16

V66

.000
.424
.507
.377
.009

00000

O

.072

0.228

WFR1

V71
0.719
-0.348

AVERAGE OFF-DIAGONAL ABSOLUTE RESIDUAL

MAXIMUM LIKELIHOOD SOLUTION

BUR12

V62

0.000

-0.400

-0.129

-0.584

-0.180

0.061

0.321

-0.459

-0.126

0.720

-0.393

BUR17

V67

.000
.294
.194
.053

0000

O

.048

-0.267

WFR2
V72

0.006

AVERAGE ABSOLUTE RESIDUAL

BUR13

V63

.000

.466

.251

.080

.747

.384

.478

.146

.853

.321

BUR18

O

V68

.000
.155
.001
.753

.797

BUR14

V64

.000
.313

.369

.027

.079

.515

.012

.585

.174

BUR19

(NORMAL DISTRIBUTION THEORY)

-220-

V69

.000
.084

.365

.174

STANDARDIZED RESIDUAL MATRIX:

PRES
VS
PRE1
PRE2
PRE3
PRE4
PRES
0.000
PRE6
0.091
PRE7
0.043
PRE8
0.088
DIR1
0.043
DIR2
0.087
DIR3
0.078
DIR4
0.018
DIRS
0.016
DIR6
0.039
DIR7
0.029
DIR8
0.041
DIR9
0.055
DIR10
0.088
DIR11
0.025
DIR12
0.090
DIR13
0.034
RESI
0.040
RESZ
0.060
RESB
0.034
RES4
0.023
RESS
0.033
RES6
0.018

V1

V2

V3

V4

V5

V6

V7

V8

V9

V10

V11

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

PRE1

V1

.000

.045

.010

.043

.140

.006

.007

.026

.002

.043

.027

.064

.015

.068

.147

.026

.023

.014

.008

.063

.001

.040

.106

.079

.025

.011

.075

PRE2

V2

.000

.079

.016

.071

.091

.100

.012

.018

.061

.028

.065

.034

.044

.012

.041

.055

.024

.065

.033

.057

.027

.004

.052

.024

.014

.005

PRE3

V3

.000

.016

.122

.090

.097

.080

.012

.086

.048

.003

.030

.047

.027

.003

.011

.027

.119

.078

.087

.030

.049

.038

.008

.014

.086

P

O

RE4

V4

.000
.034

.025

.036

.057

.033

.059

.005

.053

.091

.000

.038

.012

.053

.069

.024

.011

.021

.018

.038

.053

.011

.029

.026

 

RES7
.039
RES8
.030
RES9
.120
RESlO
.099
RESll
.028
WTE1
.007
WTE2
.054
WTE3
.010
WTE4
.031
WTES
.048
WTE6
.044
WTE7
.028
WTE8
.020
WTE9
.027
WTE10
.048
WTE11
.050
VAR1
.061
VAR2
.096
VAR3
.038
VAR4
.037
VARS
.033
VAR6
.053
VAR7
.056
BUR1
.032
BUR2
.099
BUR3
.060
BUR4
.078
BURS
.051

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

.019

.012

.084

.004

.060

.077

.026

.000

.012

.020

.024

.035

.037

.016

.088

.071

.032

.011

.012

.032

.011

.005

.022

.084

.072

.089

.018

.031

-222-

.075

.005

.047

.045

.060

.004

.076

.060

.006

.018

.017

.018

.046

.004

.044

.046

.030

.110

.062

.039

.018

.017

.032

.021

.022

.007

.064

.024

.003

.024

.092

.001

.001

.028

.007

.066

.084

.128

.094

.023

.043

.044

.006

.005

.005

.017

.019

.004

.027

.038

.064

.086

.000

.082

.042

.034

.004
.053
.147
.005
.082
.031
.087
.049
.067
.023
.013
.077
.119
.025
.033
.099
.005
.081
.041
.057
.057
.037
.025
.029
.015
.006
.072

.044

0.039

0.012

0.028

0.046

0.018

0.041

0.133

0.117

0.089

0.082

0.038

0.062

0.052

0.021

0.002

0.027

0.096

DIR2

V10

0.000

0.002

0.132

0.032

0.071

0.045

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

PRE6
PRE7
PRE8
DIR1
DIR2
DIR3
DIR4
DIR5
DIR6

DIR7

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V10

V11

V12

V13

V14

V15

0.005

0.005

0.056

0.000

0.115

0.049

-0.077

0.024

0.012

-0.032

0.011

0.028

0.015

0.117

-0.065

-0.031

-0.062

PRE6

V6

.000
.012
.083
.051
.036

00000

O

.058

-0.084

0.020

-0.035

-0.018

-0.038
0.058
0.086
0.061

-0.012

-0.041
0.018

-0.014

-0.061

-0.082
0.013
0.004
0.018
0.012
0.031
0.018

0.030

PRE7
V7
0.000
0.022
0.023
0.020
-0.017
0.099
-0.103

-0.030

0.049

~223-

-0.021

-0.069

0.123

-0.007

0.008

-0.042

0.112

0.013

0.038

-0.033

0.077

0.026

0.101

-0.092

0.005

-0.032

-0.080

PRE8

V8

0.000
0.039
-0.059
0.003
-0.052
-0.042

0.038

0.022

-0.064

-0.026

0.035

0.016

0.039

0.049

0.093

0.000

-0.081

0.009

0.052

-0.010

0.063

0.015

-0.048

0.035

-0.067

DIR1

V9

0.000
0.013

-0.041

0.021

0.048

0.026

0.015

DIR8
.061
DIR9
.023
DIR10
.018
DIR11
.003
DIR12
.091
DIR13
.081
RESl
.023
RESZ
.001
RES3
.026
RES4
.030
RESS
.080
RES6
.088
RES7
.015
RESS
.147
RES9
.011
RESlO
.020
RESll
.037
WTE1
.008
WTE2
.146
WTE3
.066
WTE4
.055
WTES
.077
WTE6
.015
WTE7
.031
WTE8
.005
WTE9
.065
WTE10
.010
WTE11
.021

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

.034

.030

.072

.015

.005

.016

.044

.084

.007

.035

.044

.050

.005

.044

.037

.050

.125

.077

.064

.009

.067

.032

.054

.020

.002

.014

.030

.045

~XM-

.055

.025

.021

.062

.077

.042

.071

.040

.010

.085

.024

.039

.044

.104

.031

.052

.022

.125

.067

.055

.086

.057

.022

.003

.087

.086

.016

.002

.108

.006

.074

.010

.011

.009

.037

.049

.028

.010

.055

.061

.057

.067

.090

.062

.048

.034

.045

.017

.034

.035

.079

.022

.001

.041

.086

.028

.003

.064

.073

.015

.013

.063

.068

.097

.000

.025

.018

.010

.019

.041

.007

.047

.009

.033

.030

.015

.034

.009

.046

.077

.008

.005

.009

.014

.023

.041

.066

.088

.030

.071

.051

.001

.011

.075

.010

.013

.056

.036

.018

.015

.031

.074

.019

.047

.013

.046

.000

.022

.037

.102

.006

.028

VAR1

VAR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

.062

.004

.006

.001

.017

.061

.022

.009

.052

.141

.015

.032

.081

.048

.002

.054

.042

.030

.015

.033

.020

.029

.079

.043

.079

.029

.019

.058

—225-

.118

.142

.073

.051

.055

.031

.009

.033

.042

.002

.064

.052

.036

.064

.043

.054

.052

.068

.002

.024

.110

.063

.094

.068

.020

.015

.057

.026

.006

.008

.004

.055

.047

.038

.032

.084

.019

.073

.013

.028

.010

.017

.076

.028

.021

.012

.059

.004

.011

.026

.072

.065

.085

.030

.019

.034

.017

.023

.036

.089

.050

.012

.008

.092

.104

.017

.043

.010

.033

.035

.045

.028

.002

.003

.040

.021

.031

.078

.054

.007

.006

.024

.063

.022

 

WFR2
0.071
DIR7
V15
DIR3
DIR4
DIRS
DIR6
DIR7
0.000
DIR8
0.105
DIR9
0.054
DIR10
0.043
DIR11
0.053
DIR12
0.031
DIR13
0.119
RESl
0.045
RESZ
0.015
RESB
0.035
RES4
0.011
RESS
0.018
RESG
0.001
RES7
0.034
RESS
0.009
RES9
0.019
RESlO
0.011
RESll
0.068
WTE1
0.047
WTE2
0.022
WTE3
0.174
WTE4
0.071

V72

V11

V12

V13

V14

V15

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

0

.050

DIR3

V11

.000

.022

.026

.031

.021

.003

.081

.122

.020

.009

.017

.054

.031

.009

.095

.000

.040

.031

.027

.055

.050

.119

.086

.072

.064

.054

0.

023

DIR4

V12

-226-

.000

.020

.020

.029

.026

.033

.031

.031

.017

.050

.003

.044

.037

.063

.131

.031

.051

.038

.013

.010

.081

.025

.006

.036

.003

0

.572

DIRS

-0.
-0.

V13

.000

054

012

.113

.068

.010

.037

.020

.028

.020

.073

.049

.003

.001

.026

.144

.023

.050

.001

.058

.040

.026

.023

.039

0

D

.019

IR6

V14

.000
.004

.051

.038

.026

.076

.033

.056

.019

.065

.048

.052

.050

.002

.086

.042

.049

.033

.065

.047

.007

.007

.028

 

WTES
.010
WTE6
.051
WTE7
.054
WTE8
.026
WTE9
.062
WTE10
.005
WTE11
.012
VAR1
.044
VAR2
.001
VAR3
.053
VAR4
.009
VARS
.053
VAR6
.048
VAR7
.024
BUR1
.071
BUR2
.014
BUR3
.007
BUR4
.067
BURS
.009
BUR6
.018
BUR7
.019
BUR8
.060
BUR9
.009
BUR10
.017
BUR11
.012
BUR12
.046
BUR13
.070
BUR14
.095

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

.036

.030

.117

.020

.039

.033

.037

.005

.056

.042

.038

.017

.005

.032

.013

.045

.079

.069

.085

.061

.006

.013

.035

.044

.015

.034

.005

.001

-227-

.036

.021

.104

.022

.069

.007

.021

.064

.002

.027

.042

.032

.050

.043

.064

.090

.034

.060

.050

.100

.002

.034

.073

.019

.041

.025

.065

.004

.020

.096

.039

.026

.012

.042

.030

.079

.068

.051

.005

.040

.118

.003

.021

.018

.037

.007

.003

.008

.035

.036

.026

.054

.043

.038

.031

.018

.007

.061

.004

.026

.030

.052

.065

.006

.005

.028

.016

.082

.100

.003

.037

.047

.096

.027

.002

.014

.029

.074

.117

.023

.017

.018

.001

.076

 

BUR15
0.003
BUR16
0.045
BUR17
0.007
BUR18
0.069
BUR19
0.157
BUR20
0.035
WFR1
0.010
WFR2
0.043
DIR12
V20
DIR8
DIR9
DIR10
DIR11
DIR12
0.000
DIR13
0.030
RESl
0.047
RESZ
0.010
RES3
0.088
RES4
0.116
RESS
0.004
RES6
0.026
RES7
0.018
RES8
0.031
RES9
0.051
RESlO
0.012
RESll
0.014
WTE1
0.047
WTE2

0.067

V65

V66

V67

V68

V69

V70

V71

V72

V16

V17

V18

V19

V20

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

-0.073

-0.041

-0.087

—0.035

0.014

0.016

0.019

0.059

DIR8

V16

0.000

0.024

-0.080

-0.066

-0.091

-0.021

-0.021

-0.004

0.089

0.041

0.010

0.009

0.007

0.056

0.045

0.035

-0.014

0.047

0.054

0.092
0.011
0.038
-0.027
0.059
-0.082
0.059

-0.005

DIR9
v17
0.000
0.041
-0.078
0.051
-o.029
40.033
0.011
0.018
0.009
0.001
0.010
0.072
-o.019
-0.065
0.038
0.017
0.024

0.001

-0.034

0.041

-0.030

0.092

0.008

0.086

0.170

-0.089

DIR10

V18

0.000

-0.078

0.033

-0.088

-0.050

-0.007

—0.023

-0.032

-0.002

-0.035

0.006

-0.012

-0.023

0.042

-0.042

-0.010

0.003

—0.022

-0.033

-0.033

-0.052

0.024

-0.037

-0.016

0.462

DIR11

V19

0.000
0.006

-0.096

-0.018

-0.013

0.037

-0.001

0.019

0.051

-0.097

-0.003

0.057

-0.035

-0.136

0.001

-0.033

 

 

WTE3
.092
WTE4
.009
WTES
.025
WTE6
.037
WTE7
.112
WTE8
.041
WTE9
.113
WTE10
.089
WTE11
.020
VAR1
.012
VAR2
.017
VAR3
.017
VAR4
.056
VARS
.008
VAR6
.048
VAR7
.005
BUR1
.002
BUR2
.039
BUR3
.078
BUR4
.010
BURS
.110
BUR6
.027
BUR7
.007
BUR8
.016
BUR9
.022
BUR10
.059
BUR11
.031
BUR12
.012

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

.014

.023

.051

.056

.029

.003

.019

.085

.036

.036

.024

.031

.002

.068

.015

.005

.038

.064

.001

.019

.056

.064

.015

.022

.003

.018

.043

.001

-229-

.067

.012

.085

.026

.055

.072

.034

.012

.035

.036

.052

.083

.016

.049

.074

.008

.001

.034

.029

.010

.030

.068

.020

.012

.007

.064

.009

.025

.061

.034

.012

.021

.037

.021

.107

.007

.014

.061

.016

.011

.150

.048

.015

.020

.015

.049

.012

.003

.035

.040

.032

.016

.024

.054

.102

.059

.046

.006

.054

.076

.004

.017

.050

.037

.033

.029

.012

.062

.049

.053

.015

.004

.097

.084

.014

.079

.075

.062

.063

.025

.058

.014

.028

.085

0.016

0.013

0.028

0.066

0.013

0.006

0.019

0.019

0.037

0.062

RES4

V25

0.000

0.017

0.021

0.038

0.059

0.031

0.025

0.003

0.030

0.055

0.073

0.032

0.029

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

DIR13

RESl

RESZ

RES3

RES4

RESS

RES6

RES7

RESS

RES9

RESlO

RESll

WTE1

WTE2

WTE3

WTE4

WTES

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V21

V22

V23

V24

V25

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

0.020
0.009
-0.022
0.048
0.068
0.056
-0.009
0.071
0.014

-0.035

DIR13
V21
0.003
0.035
0.072
0.001
0.001
-0.023
-0.026
-0.029
0.063
0.000
-0.032
0.054
0.032
0.052
-0.033

0.121

0.110

-0.050
0.076
0.018
0.025
0.031

~0.108

-0.032
0.023

-0.064

-0.091

RESI
V22
0.000
0.107
0.004
0.000
-0.010
-0.028
0.085
0.042
0.023
0.074
-0.016
-0.093
-0.050
0.046

0.012

0.071

-230-

0.027

0.077

-0.149

-0.033

-0.026

-0.021

-0.085

-0.016

0.010

0.041

RES2

V23

0.000

0.036

-0.008

0.008

0.053

-0.063

0.099

-0.040

-0.034

~0.029

-0.043

0.044

0.000

0.027

0.056

0.026

0.023

0.015

-0.002

-0.016

-0.030

—0.053

-0.079

-0.037

-0.075

RES3

V24

0.000
-0.053

-0.020

0.050

-0.034

-0.024

0.027

-0.058

-0.012

0.018

0.013

-0.107

0.018

-0.063

.002

.055

.034

.020

.085

.014

.086

.008

.009

.022

.014

.029

.017

.008

.052

.077

.030

.033

.004

.018

.020

.100

.043

.004

.087

.082

.005

.019

WTE6

WTE7

WTE8

WTE9

WTE10

WTE11

VAR1

VAR2

VAR3

VAR4

VARS

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

V38

V39

V40

V41

V42

V43

V44

V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

.128

.063

.027

.011

.033

.013

.001

.038

.007

.032

.037

.001

.045

.042

.079

.024

.016

.000

.003

.055

.067

.038

.120

.015

.053

.077

.049

.047

-23l-

.077

.034

.044

.064

.055

.050

.002

.090

.044

.042

.046

.008

.035

.041

.018

.013

.011

.022

.016

.117

.002

.014

.050

.077

.017

.056

.029

.058

.080

.066

.051

.018

.045

.066

.013

.078

.029

.027

.052

.001

.030

.108

.040

.002

.101

.009

.003

.053

.030

.073

.009

.006

.046

.027

.078

.017

.143

.075

.041

.041

.011

.111

.038

.066

.041

.125

.117

.054

.047

.001

.003

.014

.081

.012

.070

.075

.027

.022

.039

.047

.014

.035

.020

.028

RES9

V30

BUR16
.009
BUR17
.017
BUR18
.044
BUR19
.009
BUR20
.029
WFR1
.009
WFR2
.028
RESS
RES6
RES7
RESB
RES9
.000
RESlO
.045
RESll
.008
WTE1
.013
WTE2
.031
WTE3
.029
WTE4
.011
WTES
.056
WTE6
.035
WTE7
.024
WTE8
.005
WTE9
.002
WTE10
.007
WTE11
.058
VAR1
.021
VAR2
.015

V66

V67

V68

V69

V70

V71

V72

V26

V27

V28

V29

V30

V31

V32

V33

V34

V35

V36

V37

V38

V39

V40

V41

V42

V43

V44

V45

0.037

-0.024

-0.063

-0.004

-0.004

-0.067

0.058

RESS

V26

0.000

-0.016

0.058

-0.002

0.047

-0.029

0.018

-0.003

0.048

—0.017

0.026

-0.032

0.066

0.084

0.027

0.003

0.058

0.001

0.122

0.009

-0.006

0.013

~0.012

-0.037

0.046

0.042

-0.033

RES6

V27

0.000

-0.024

-0.047

0.024

0.004

—0.008

0.014

-0.041

0.009

-0.141

-0.025

0.034

-0.003

0.036

0.076

-0.092

0.049

-0.011

0.013

-232-

0.000

-0.039

0.006

-0.076

0.094

0.056

-0.011

RES7

V28

0.000

-0.089

0.032

0.060

0.035

-0.058

-0.079

-0.018

-0.039

0.010

0.053

0.032

0.005

-0.031

-0.052

0.008

0.030

-0.086

-0.020

0.047

-0.070

-0.092

0.048

-0.006

0.079

RESB

V29

0.000
-0.066

-0.078

-0.031

0.021

0.011

0.107

0.031

0.024

0.079

0.028

0.023

0.003

0.001

0.031

-0.019

0.010

VAR3 V46 0.015 0.010 0.017 -0.050

0.027
VAR4 V47 0.078 0.008 -o.042 0.049 -
0.024
VARS V48 0.000 0.029 0.006 0.005 -
0.053
VAR6 V49 0.053 0.071 0.140 0.096 —
0.009
VAR7 vso -o.oos -o.04o -0.064 —o.o75 -
0.040
BUR1 vs1 0.010 0.072 —0.038 0.017 -
0.039
BUR2 vs2 -o.017 0.042 -o.020 0.054
0.065
BUR3 vs3 -0.038 0.012 0.020 -o.049 -
0.044
BUR4 vs4 -o.073 -o 019 -o.020 0.042 -
0.008
BURS vss —o.015 0.025 —0.058 -o.01o -
0.045
BUR6 vss 0.018 0.053 -0.062 0.006 —
0.090
BUR7 vs7 -0.036 -0.069 -o.043 0.138
0.024 .
BUR8 vse -o.045 —o.054 0.051 -o.o31
0.049
BUR9 V59 0.040 -o.oos 0.021 -o.001
0.066
BUR10 V60 0.083 0.045 0.035 -o.o37
0.059
BUR11 V61 -o.o37 -0.038 0.018 -o.009
0.019
BUR12 V62 0.023 -o.o3o -o.029 0.093
0.140
BUR13 V63 -0.082 0.054 -o.135 0.004
0.032
BUR14 V64 0.027 0.045 0.007 -o.o37 -
0.011
BUR15 V65 —o.01o -o.021 0.045 —O.148
0.040
BUR16 V66 0.015 -o.021 -o.020 0.038 -
0.022
BUR17 V67 0.006 0.015 0.017 -o.041
0.020
BUR18 V68 -0.066 0.040 -O.116 0.054
0.020
BUR19 V69 0.078 -o.009 -o.o33 -o.015 -
0.029
BUR20 V70 0.062 0.012 0.006 0.026
0.010
WFR1 V71 -o.o39 0.031 -0.056 0.049 -
0.011
WFR2 V72 0.027 -o.011 0.010 -o.007 -
0.048

-233-

 

 

RESlO RESll WTE1 WTE2

WTE3
V31 V32 V33 V34
V35
RESlO V31 0.000
RESll V32 0.021 -0.002
WTEl V33 0.008 0.067 0.000
WTE2 V34 -0.011 0.064 0.049 0.000
WTE3 V35 -0.004 -0.025 -0.105 -0.036
0.000
WTE4 V36 0.054 0.065 0.020 -0.008
0.026
WTES V37 -0.102 -0.001 0.023 -0.010
0.007
WTE6 V38 -0.069 -0.047 -0.014 -0.002
0.015
WTE7 V39 0.045 0.017 -0.003 0.092
0.036
WTE8 V40 -0.068 -0.022 0.101 0.055
0.026
WTE9 V41 0.007 -0.027 -0.053 -0.008
0.059
WTE10 V42 0.019 -0.079 -0.049 —0.043
0.032
WTE11 V43 0.005 -0.025 0.016 -0.007
0.005
VAR1 V44 -0.013 0.000 -0.026 0.028
0.036 -
VAR2 V45 0.065 0.015 -0.037 0.026
0.051
VAR3 V46 0.020 -0.020 -0.060 —0.019
0.040
VAR4 V47 -0.028 -0.015 0.064 -0.010
0.036
VARS V48 -0.017 -0.010 0.047 0.023
0.010
VAR6 V49 0.002 -0.049 0.011 -0.034
0.121
VAR7 V50 0.013 -0.011 -0.017 —0.055
0.033
BUR1 V51 -0.029 0.049 -0.034 -0.002
0.012
BUR2 V52 0.011 0.038 -0.056 0.109
0.040
BUR3 V53 -0.065 0.055 -0.020 -0.049
0.071
BUR4 V54 -0.079 -0.127 -0.014 0.052
0.022
BURS VSS 0.077 -0.059 -0.105 -0.025
0.027
BUR6 V56 -0.038 -0.054 0.038 -0.063
0.036
BUR7 V57 -0.076 0.048 0.046 0.028
0.051
BUR8 V58 0.082 -0.012 0.023 -0.004
0.058

-234-

BUR9
0.001
BUR10
0.056
BUR11
0.005
BUR12
0.028
BUR13
0.030
BUR14
0.000
BUR15
0.043
BUR16
0.026
BUR17
0.003
BUR18
0.049
BUR19
0.018
BUR20
0.125
WFR1
0.038
WFR2
0.015

WTE8

V40

WTE4

WTES

WTE6

WTE7

WTE8
0.000

WTE9
0.009

WTE10
0.011
WTE11

0.006

VAR1
0.042

VAR2
0.020

VAR3
0.086

VAR4
0.055

VARS
0.006

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V36
V37
V38
V39
V40
V41
V42
V43
V44
V45
V46

V47

V48

-0.027
-0.014
0.016
-0.026
0.056
0.026
-0.042
0.082
-0.004
-0.008
-0.031
0.084
-0.068

0.029

WTE4
V36
0.000
0.069
0.029
0.018
0.006
-0.008
-0.035
0.017
-0.013
0.018
—0.057

-0.066

0.019

-0.004
-0.005
0.073
0.027
0.008
-0.016
-0.018
0.023
0.040
0.047
-0.099
0.039
-0.040

-0.011

WTES

V37
0.000
0.047
-0.050
0.072
0.083
-0.059
0.025
0.023
-0.043
0.006

0.115

-0.006

-235-

0.057

-0.016

0.037

0.050

-0.071

-0.002

-0.037

-0.099

-0.048

-0.014

0.037

0.007

0.010

-0.007

WTE6

V38

0.000
0.001
-0.018
-0.011
-0.035
-0.011
-0.040
0.029
-0.037

-0.020

-0.047

-0.022

0.003

-0.081

-0.021

0.057

0.029

0.023

0.035

-0.045

0.083

-0.010

0.079

-0.023

0.001

WTE7

V39

0.000
0.020

-0.088

-0.031

-0.004

-0.015

0.085

—0.106

0.041

-0.086

0.096

0.059

0.035

0.032

0.090

0.057

0.094

0.083

0.022

0.030

0.084

0.007

0.066

0.116

0.032

0.075

0.109

0.021

0.042

0.001

0.013

0.078

0.041

0.005

VAR2

V45

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

BUR10

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

WTE9

WTE10
WTE11

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V41
V42
V43

-0.042

0.094

0.005

0.078

-0.017

0.002

0.037

-0.076

0.053

0.038

0.007

-0.072

-0.043

-0.056

-0.019

0.001

-0.063

-0.011

-0.098

-0.045

-0.033

-0.078

-0.026

0.001

WTE9
V41
0.000

0.072
-0.020

-0.045

-0.078

-0.052

0.010

-0.008

-0.037

-0.012

-0.003

0.002

0.000

-0.067

-0.079

0.023

-0.045

0.056

-0.104

-0.019

0.002

0.107

0.020

0.010

-0.097

0.008

-0.016

WTE10

V42

0.000
0.000

-0.098

0.012

-0.015

-0.017

-0.080

0.011

-0.033

0.020

0.010

0.071

0.030

-0.025

0.059

-0.014

-0.033

0.040

0.003

-0.006

-0.037

0.008

-0.010

0.086

0.024

0.085

WTE11

V43

-0.003

0.015

0.023

—0.003

0.013

0.072

0.014

0.075

0.013

0.040

—0.048

0.029

-0.006

0.006

0.017

0.045

-0.053

0.029

0.093

0.048

0.089

-0.077

0.026

0.021

-0.013

VAR1

V44

VAR1
VAR2
.000
VAR3
.032
VAR4
.075
VARS
.008
VAR6
.009
VAR7
.053
BUR1
.003
BUR2
.046
BUR3
.008
BUR4
.063
BURS
.026
BUR6
.033
BUR7
.OSS
BUR8
.024
BUR9
.013
BUR10
.031
BUR11
.054
BUR12
.038
BUR13
.058
BUR14
.035
BUR15
.018
BUR16
.006
BUR17
.044
BUR18
.013
BUR19
.106
BUR20
.019
WFR1
.024
WFR2
.014

V44
V45

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

.085
.008

.015

.019

.022

.102

.079

.027

.081

.033

.056

.002

.014

.064

.043

.004

.029

.043

.029

.011

.045

.060

.030

.024

.027

.031

.036

.029

.024

-237-

.014
.064

.023

.011

.099

.038

.070

.069

.007

.004

.059

.090

.092

.035

.066

.088

.005

.004

.014

.013

.157

.020

.004

.011

.022

.040

.106

.018

.015

.068
.026

.039

.033

.060

.026

.023

.001

.049

.001

.042

.054

.033

.055

.011

.043

.001

.004

.107

.034

.062

.063

.006

.003

.081

.072

.053

.065

.026

.000
.096

.019

.020

.002

.027

.018

.081

.035

.099

.085

.070

.038

.017

.042

.043

.012

.063

.073

.091

.092

.008

.009

.023

.039

.061

.044

.024

.011

 

VAR7
V50
VAR3
VAR4
VARS
VAR6
VAR7
0.000
BUR1
0.070
BUR2
0.093
BUR3
0.089
BUR4
0.022
BURS
0.049
BUR6
0.064
BUR7
0.074
BUR8
0.055
BUR9
0.007
BUR10
0.017
BUR11
0.133
BUR12
0.046
BUR13
0.030
BUR14
0.038
BUR15
0.068
BUR16
0.096
BUR17
0.041
BUR18
0.069
BUR19
0.028
BUR20
0.030
WFR1
0.067
WFR2
0.003

V46

V47

V48

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

V60

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

VAR3

V46

.000

.041

.004

.080

.037

.003

.062

.058

.005

.006

.003

.068

.028

.083

.022

.063

.112

.042

.032

.064

.071

.025

.043

.035

.073

.046

.036

VAR4

V47

.000

.015

.034

.006

.023

.026

.073

.040

.047

.090

.090

.047

.013

.025

.072

.028

.018

.001

.038

.083

.076

.029

.003

.012

.023

.007

VARS

V48

.000

.004

.015

.005

.052

.028

.068

.003

.039

.034

.034

.085

.032

.055

.003

.064

.026

.130

.027

.044

.018

.007

.023

.025

.010

VAR6

V49

.000
.021

.010

.029

.032

.105

.023

.025

.053

.024

.017

.070

.092

.042

.065

.027

.084

.061

.013

.035

.017

.000

.039

.034

 

J’

 

BUR1 BUR2 BUR3 BUR4
BURS
vs1 vs2 vs3 vs4
vss
BUR1 vs1 0.000
BUR2 vs2 0.055 0.000
BUR3 vs3 0.061 -o.011 0.000
BUR4 vs4 -o 011 -0.006 0.033 0.000
BURS vss -o.oo3 0.093 0.066 0.031
0.000
BUR6 V56 -o.107 -o.07o 0.024 0.085 -
0.006
BUR7 vs7 -o.055 -01094 -o.04o -o.o34 -
0.029
BUR8 vse 0.079 0 009 —0.062 -o.024 -
0.102
BUR9 vs9 —0.006 —0.056 0.033 0.013 -
0.032
BUR10 V60 0.027 -o.oo2 -o.052 —0.067
0.021
BUR11 V61 0.017 0.012 0.030 -0.o32 -
0.010
BUR12 V62 0.090 0.019 -0.008 -0.062
0.045
BUR13 V63 —o.005 0.002 -0.016 0.000
0.062
BUR14 V64 0.054 -0.08O 0.084 0.127 -
0.001
BUR15 V65 0.042 0.018 0.005 0.002
0.058
BUR16 V66 0.033 -o.oos —o.01o -0.018 -
0.041
BUR17 V67 0.052 -o.o31 -0.028 0.041
0.042
BUR18 V68 -o.054 -o.009 -o.002 0.029 —
0.059
BUR19 V69 0.016 0.028 -o.042 0.047 -
0.016
BUR20 V70 0.007 -o.012 -o.011 0.008
0.005
WFR1 V71 0.011 -0 007 0.045 0.005
0.067
WFR2 V72 -0.016 -o.017 -o.119 0.041
0.013
BUR6 BUR7 BUR8 BUR9
BUR10
vss V57 vss V59
V60
BUR6 vss 0.000
BUR7 vs7 -o 057 0.000
BUR8 vse 0.043 -o.012 0.000
BUR9 vs9 0.024 -o.012 -o.029 0.000
BUR10 V60 0.017 —o.052 0.057 -0.018
0.000

-239-

 

 

0.026

0.074

0.024

0.079

0.080

0.018

0.007

0.059

0.023

0.006

0.037

0.036

BUR15

V65

0.000

0.054

0.066

0.011

0.006

0.002

0.084

0.021

BUR20

V70

BUR11

BUR12

BUR13

BUR14

BUR15

BUR16

BUR17

BUR18

BUR19

BUR20

WFR1

WFR2

BUR11
BUR12
BUR13
BUR14
BUR15
BUR16
BUR17
BUR18
BUR19

BUR20

WFR1

WFR2

BUR16

V61

V62

V63

V64

V65

V66

V67

V68

V69

V70

V71

V72

V61
V62
V63
V64
V65
V66
V67
V68
V69
V70
V71

V72

V66

0.027

-0.057

-0.064

0.059

0.074

-0.057

-0.064

0.016

-0.055

0.004

0.011

0.020

BUR11

V61

.000
.090
.047
.003
.048

00000

O

.033

-0.015

0.018

0.023

0.001

—0.036

0.044

BUR16

V66

0.000

0.056

0.064

0.000

0.015

-0.078

-0.072

—0.054

-0.043

0.023

0.001

0.000

0.054

BUR12

V62

0.000

-0.050

-0.016

-0.074

-0.023

0.007

0.039

-0.057

-0.016

0.088

-0.048

BUR17

V67

-240-

-0.111

-0.015

0.036

0.047

-0.031

0.043

-0.005

-0.073

-0.022

0.011

0.053

0.037

BUR13

V63

0.000

-0.058

0.032

0.010

0.093

0.048

0.060

-0.018

0.106

-0.040

BUR18

V68

0.019

0.102

-0.070

0.033

-0.032

-0.060

0.095

-0.043

-0.044

0.000

0.018

0.007

BUR14

V64

0.000
—0.039

-0.046

-0.003

-0.010

-0.063

-0.001

-0.071

-0.021

BUR19

V69

BUR17
BUR18
BUR19
BUR20
0.004
WFR1
0.029
WFR2
0.042

WFR1
WFR2

0.0389

AVERAGE OFF—DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL

0.0399

LARGEST STANDARDIZED

Z
O

|._.|
Okomxlmmbfiwa-‘l

V67
V68
V69
V70

V71

V72

V71
V72

AVERAGE ABSOLUTE STANDARDIZED RESIDUAL

.053
.063
.048
.001

.009

.029

WFR1
V71

0.
-0.

087
042

.000
.036
.024
.006

.006

.032

WFR2
V72

0.

RESIDUALS:
PARAMETER ESTIMATE
V72, V8 0.572
V72, V14 0.462
V35, V15 0.174
V71, V13 0.170
V64, V42 0.157
V69, V15 0.157
V47, V18 -0.150
V65, V18 -0 149
V65, V29 -0 148
V15, V1 0 147

MAXIMUM LIKELIHOOD SOLUTION

001

NO.
11
12
13
14
15
16
17
18
19
20

0.000
.019
0.000

O

0.091

-0.097

PARAMETER

0.000
-0.010

0.045

0.022

ESTIMATE

(NORMAL DISTRIBUTION THEORY)

DISTRIBUTION OF STANDARDIZED RESIDUALS

FREQ PERCENT
975-

*X’I'i-

it— 1'

16-41-36!-

* #-

-24l-

RANGE

. fl

 

! * * l 3 -0.3 — —0.4

0 0.00%

! * * l 4 —0.2 - -0.3
0 0.00%

650- * * - 5 -0.1 - —O.2

54 2.05%

! * * l 6 0.0 — —0.1
1211 46.08%

! * * l 7 0.1 — 0.0
1305 49.66%

I * * ' 8 0.2 - 0 1
56 2.13%

1 * * ' 9 0.3 - 0 2
O 0.00%

325- ** - A 04- 03

O 0.00%

! * * ' B 0.5 — O 4
1 0.04%

! * * ' C ++ - 0 S
1 0.04%

l ** l ______________________

1 * * * * ' TOTAL
2628 100.00%

1 2 3 4 5 6 7 8 9 A B C EACH "*" REPRESENTS 65

RESIDUALS

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)

*** WARNING *** TEST RESULTS MAY NOT BE APPROPRIATE DUE TO CONDITION
CODE

GOODNESS OF FIT SUMMARY FOR METHOD = ML

INDEPENDENCE MODEL CHI—SQUARE = 3399.033 ON 2556 DEGREES
OF FREEDOM
INDEPENDENCE AIC = -1712.967 INDEPENDENCE CAIC = -14471.151
MODEL AIC = -1939.47S MODEL CAIC = -14338.273
CHI-SQUARE = 3028.525 BASED ON 2484 DEGREES OF FREEDOM
PROBABILITY VALUE FOR THE CHI-SQUARE STATISTIC IS 0.00000

THE NORMAL THEORY RLS CHI-SQUARE FOR THIS ML SOLUTION IS
2755.168.

FIT INDICES

BENTLER-BONETT NORMED FIT INDEX = 0.109
BENTLER-BONETT NON-NORMED FIT INDEX = 0.335
COMPARATIVE FIT INDEX (CFI) = 0.354
BOLLEN'S (IFI) FIT INDEX = 0.405
MCDONALD'S (MFI) FIT INDEX = 0.506

-242-

JORESKOG-SORBOM'S GFI FIT INDEX = 0.839
JORESKOG-SORBOM'S AGFI FIT INDEX = 0.830
ROOT MEAN-SQUARE RESIDUAL (RMR) = 0.423
STANDARDIZED RMR - 0.051

ROOT MEAN-SQUARE ERROR OF APPROXIMATION (RMSEA) =

0.023
90% CONFIDENCE INTERVAL OF RMSEA ( 0.020, 0.026)
RELIABILITY COEFFICIENTS
CRONBACH'S ALPHA = 0.058
ITERATIVE SUMMARY
PARAMETER
ITERATION ABS CHANGE ALPHA FUNCTION
1 7.926075 1.00000 20.95421
2 7.547451 1.00000 10.74106
3 0.857517 1.00000 8.80036
4 0.099451 1.00000 8.32817
5 0.100135 0.50000 8.00445
6 0.048024 1.00000 7.80307
7 0.040574 1.00000 7.66367
8 0.040231 0.50000 7.62924
9 0.047112 0.50000 7.62249
10 0.099768 0.20282 7.61732
11 0.153860 0.09412 7.61072
12 0.197208 0.08483 7.60591
13 0.255429 0.07994 7.60271
14 0.339163 0.07819 7.60138
15 0.470522 0.03517 7.59930
16 0.558555 0.03173 7.59760
17 0.661758 0.02944 7.59621
18 0.786112 0.02805 7.59514
19 0.941073 0.02740 7.59443
20 1.142138 0.02740 7.59419
21 1.416052 0.01240 7.59351
22 1.577285 0.01143 7.59290
23 1.752013 0.01065 7.59236
24 1.942707 0.01003 7.59188
25 2.152681 0.00955 7.59146
26 2.386397 0.00920 7.59109
27 2.649861 0.00897 7.59077
28 2.951232 0.00884 7.59053
29 3.301947 0.00882 7.59036
30 3.718169 0.00890 7.59029

'k********************‘k‘k************************************************

***************‘k‘k‘k‘k****************************

** NOTE: DO NOT TRUST THIS OUTPUT. ITERATIVE PROCESS HAS NOT
CONVERGED. MAXIMUM NUMBER OF ITERATIONS WAS REACHED.3 **

 

3 In Maximum Likelihood Estimation (MLE) method, such warning is typical. The results shown provide
the best estimate based on the amount of convergence that is reached. When the iterative process does not

-243-

**

**

** 30 ITERATIONS HAVE BEEN COMPLETED AND THE PROGRAM STOPPED. CHECK
PARAMETER IDENTIFICATION. **

** IF IDENTIFIED, USE THE UPDATED PARAMETERS TO START AGAIN.

**
*‘k
**

** SUBSEQUENT METHODS WILL NOT BE COMPUTED DUE TO ITERATIVE FAILURE.

**

***************‘k‘k****************************‘k‘k‘k‘k**********************

***********************************************

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)

MEASUREMENT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

PRE8 =V8 = 1.000 F1 + 1.000 E8
DIR13 =V21 = 1.000 F2 + 1.000 E21
RESll =V32 = 1.000 F3 + 1.000 E32
WTE11 =V43 = 1.000 F4 + 1.000 E43
VARS =V48 = 1.000 F5 + 1.000 E48
BUR20 =V70 = 1.000 F6 + 1.000 E70
WFR1 =V71 = 1.000 F7 + 1.000 E71
WFR2 =V72 = .030*F7 + 1.000 E72

.074
.403

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)

CONSTRUCT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

 

converged, the results are not accurate but are sensitive to the amount of convergence that has reached. In
order to reach the full convergence, the number of iterations at the time of running the model have to be
increased, which in turn requires higher computational power (not available for this research).

-244-

F1
.005*V4

F2
.000*V12

.000*V16

.000*V20

F3

.057*V2S

.026*V29

F4
.010*V36

.080*V40

FS
.042*V47

F1

.001*V1

.004
.262

.003*V5
.005
.591
.001*V9

.007
.077

.004*V13

.051
.077

.001*V17

.007
.077

.006*V22

.044
.126

.058*V26

.044
.307

.044*V30
.043
.009
.009*V33

.048
.179

.061*V37

.047
.290

.057*V41
.047
.211

.094*V44

.079*V2

.052
.523

.003*V6
.005
.635
.000*V10

.001
.050

.000*V14

.003
.076

.000*V18

.001
.067

.062*V23

.044
.409

.023*V27

.042
.540

.112*V31
.047
.391@
.033*V34

.048
.697

.042*V38

.047
.898

.104*V42
.048
.156@

.005*V4S

-245-

.001*V3

.005
.219

.001*V7
.004
.207
.000*V11

.004
.077

.000*V15

.001
.048

.000*V19

.003
.076

.004*V24

.043
.093

.013*V28

.044
.304

.031*V3S

.044
.694

.073*V39

.047
.558

.092*V46

.005
.863

.002
.075

.004
.076

.002
.074

.043
.333

.043
.603

.048
.200

.049
.644

CONSTRUCT EQUATIONS WITH STANDARD

(CONTINUED)

1.

MAXIMUM LIKELIHOOD

F6
.064*VS4

.048*V58

.021*V62

.045*V66

F7
.212*F4

F7

MAXIMUM LIKELIHOOD

VARIANCES OF INDEPENDENT VARIABLES

STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

050
863

004*V49
049
089

SOLUTION

.029*V51

.049
.588

.096*VSS

.050
.919

.031*V59

.049
.635

.068*V63

.050
.375

.011*V67
.048
.229
.836*F1

.163
.518

.174*FS

.252
.693

SOLUTION

+

+

71.

923

.050
.105

.012*V50
.051
.242

1.

050
832

ERRORS AND TEST STATISTICS

.020*V52

.048
.417

.058*V56

.049
.168

.028*V60

.049
.572

.026*V64

.048
.547

.026*V68
.049
.543

241*F2

.632
.077

.06S*F6
.169
.384

-246-

+

+

(NORMAL DISTRIBUTION

THEORY)

.090*V53

.048
.864

.051*V57

.047
.068

.002*V61

.048
.044

.011*V65

.050
.216

.028*V69
.049
.575
.369*F3

.238

1.548

1.

000 D7

(NORMAL DISTRIBUTION THEORY)

+

.049
.865

.048
-1.334

.049
.990

.049
-.438

.050
-.902

.195
1.085

V1

V2

V3

V4

V5

V6

V7

V9

V10

V11

V12

V13

V14

v15 ‘

PRE1

PRE2

PRE3

PRE4

PRES

PRE6

PRE7

DIR1

DIR2

DIR3

DIR4

DIRS

DIR6

DIR7

14

14

14

14

14

14

14

14

14

14

14

14

14

.756*I
.620 I
.124@I

.819*I
.554 I
.124@I

.829*I
.554 I
.124@I

.135*I
.576 I
.124@I

.817*I
.624 I
.124@I

.947*I
.563 I
.124@I

.130*I
.576 I
.124@I

.610*I
.539 I
.124@I

.851*I
.627 I
.124@I

.OS3*I
.570 I
.124@I

.638*I
.612 I
.124@I

.799*I
.552 I
.124@I

.948*I
.563 I
.124@I

.705*I

-247-

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

H H H H H

V16

V17

V18

V19

V20

V22

V23

V24

V25

V26

27

V28

V29

V30

- DIR8

— DIR9

~DIR10

-DIR11

-DIR12

- RESl

- RESZ

- RES3

- RES4

- RESS

RES6

- RES7

— RESB

— RES9

14

14

14

14

14

14

8.
.587 I
.124@I

14

8.
.624 I
.124@I

14

8

.616 I
.124@I

I

.676*I
.614 I
.124@I

I

.010*I
.638 I
.124@I

I

.233*I
.583 I
.124@I

I

.114*I
.574 I
14.

124@I
I

.135*I
.576 I
.124@I

I

.102*I
.574 I
14.

124@I
I

.396*I
.594 I
.124@I

I
289*1

I
817*I

I

.254*I
.584 I
14.

124@I
I

8.677*I

14

8.
.573 I
.124@I

14

14

.614 I
.124@I

I
099*1

I

.321*I
.589 I
.124@I

I

.506*I
.602 I

-248-

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

HHHHHHHHHHHHH

14.124@I

V31 -RE810 8.070*I
.571 I
14.124@I

V33 - WTE1 7.917*I
.560 I
14.124@I

V34 - WTE2 8.090*I
.573 I
14.124@I

V35 - WTE3 9.345*I
.662 I
14.124@I

V36 - WTE4 8.049*I
.570 I
14.124@I

V37 - WTES 8.417*I
.596 I
14.124@I

V38 - WTE6 8.249*I
.584 I
14.124@I

V39 - WTE7 8.507*I
.602 I
14.124@I

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)

VARIANCES OF INDEPENDENT VARIABLES (CONTINUED)

V40 - WTE8 7.801*I
.552 I
14.124@I

WTE9 8.325*I
.589 I
14.124@I

V41

WTE10 8.118*I
.575 I
14.124@I

V42

VAR1 8.446*I
.598 I

V44

-249-

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

HHHHHHHHHHHHHH

V45

V46

V47

V49

V50

V51

V52

V53

V54

V55

V56

V57

V58

V59

VAR2

VAR3

VAR4

VAR6

VAR7

BUR1

BUR2

BUR3

BUR4

BURS

BUR6

BUR7

BUR8

BUR9

14.

14

124@I

.415*I
.596 I
14.

124@I

.380*I
.593 I
14.

124@I

.666*I
.614 I
14.

124@I

.501*I
.602 I
14.

124@I

.070*I
.571 I
14.

124@I

.178*I
.579 I
14.

124@I

.288*I
.587 I
14.

124@I

.257*I
.585 I
14.

124@I

.466*I
.599 I
14.

124@I

.767*I
.550 I
14.

124@I

.921*I
.561 I
.124@I

.611*I
.610 I
14.

124@I

.093*I
.573 I
14.

124@I

.99S*I
.566 I
14.

124@I

-250-

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

I

V60 -BUR10 8.191*I
.580 I

14.124@I

I

V61 -BUR11 8.245*I
.584 I

14.124@I

I

V62 -BUR12 8.118*I
.575 I

14.124@I

I

V63 —BUR13 7.838*I
.555 I

14.124@I

I

V64 -BUR14 8.260*I
.585 I

14.124@I

I

V65 -BUR15 7.660*I
.542 I

14.124@I

‘I

VARIANCES OF INDEPENDENT VARIABLES (CONTINUED)

V66 -BUR16 7.788*I
.551 I
14.124@I

V67 -BUR17 8.29S*I
.587 I
14.124@I

V68 -BUR18 8.195*I
.580 I
14.124@I

V69 -BUR19 8.017*I

.568 I
14.124@I

02-DEC-09 PAGE: 19 EQS Licensee:
TITLE: Model built by EQS 6 for Windows

MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
VARIANCES OF INDEPENDENT VARIABLES

STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.

~251-

HHHHHHHHHHHHHHHHHHHHHHHHH

HHHHHHHHHHHHHHHH

E8

E21

E32

E43

E48

E70

E71

E72

- PRE8

—DIR13

-RESll

-WTE11

- VARS

-BUR20

- WFR1

- WFR2

STANDARDIZED SOLUTION:
SQUARED

PRE8
.006

DIR13

.000

RESll

.027

WTE11

.033

VARS

.018

BUR20

.042

WFR1

.491

=V8

=V21

=V32 =

=V43 =

=V48 =

=V70 =

=V71 =

.076 F1

.004

.163

.182

.135

.205

.701

14

14

14

14

14

14

14

F2

F3

F4

F5

F6

F7

.418*I
.596 I
.124@I

I
.585*I
.608 I
.124@I

I
.968*I
.564 I
.124@I

I
.003*I
.567 I
.124@I

I
.821*I
.625 I
.124@I

I
.778*I
.551 I
.124@I

I
.840*I
.357 I
.410 I

I
.170*I
.578 I
.123@I

I

D7

+ .997 E8

+ l

.000

.987

.983

.991

.979

.713

-252-

E21

E32

E43

E48

E70

E71

.000*I

9.353 I

.000 I

H

HHHHHHHHHHHHHHHHHHHHHHHHHHHH

WFR2 =V72 = .020*F7 + 1.000 E72
.000
F1 =F1 = -.015*V1 + .997*V2
- .036*V5 + .039*V6
F2 =F2 = -.129*V9 — .008*V10
+ .972*V13 - .065*V14
- .135*V17 + .016*V18
F3 =F3 = -.034*V22 + .385*V23
- .3S6*V26 - .145*V27
+ .273*V30 + .680*V31
F4 =F4 = -.046*V33 + .180*V34
- .33S*V37 + .232*V38
+ .314*V41 - .S6S*V42
F5 =F5 = .675*V44 - .038*V45
- .032*V49 + .086*V50
F6 =F6 = -.l40*V51 + .099*V52
- .458*VSS + .278*V56
+ .151*V59 + .136*V60
+ .328*V63 + .130*V64
- .055*V67 - .129*V68
F7 =F7 = .901*F1 + .419*F2
+ .037*FS + .020*F6
1.000

+ + + +

.012*V3

.012*V7

.076*V11
.007*V15
.050*V19
.025*V24
.081*V28

.179*V35
.405*V39

.663*V46

.44S*V53
.254*V57
.010*V61
.051*V6S
.137*V69
.089*F3

.000 D7

BENTLER-RAYKOV CORRECTED R-SQUARED COEFFICIENTS:
SQUARED CORRELATIONS BETWEEN DEPENDENT VARIABLES AND PREDICTORS

PRE8 =V8 0.006
DIR13 =V21 0.000
RESll =V32 0.027
WTE11 =V43 0.033

VARS =V48 0.018
BUR20 =V70 0.042

WFR1 =V7l 0.491

WFR2 =V72 0.000

F1 =Fl 1.000
F2 =F2 1.000
F3 =F3 1.000
F4 =F4 1.000
FS =FS 1.000
F6 =F6 1.000
F7 =F7 1.000

today is 2009/12/02

Execution begins at 18:10:37
Execution ends at 18:10:40
Elapsed time = 3.00 seconds

-253-

.059*V4

.044*V12
.071*V16
.030*V20
.363*V25
.162*V29

.052*V36
.428*V40

.309*V47
.318*V54
.236*V58
.104*V62
.215*V66

.058*F4

APPENDIX C

Source of each question in the checklist (survey questionnaire)

Measured Variables
PRE1. Prior work is complete with desired quality.

PRE2. The pre-requisite work has been verified to meet
current work needs (dimensions, locations, etc)

PRE3. All regulatory inspections needed were conducted
and successfully passed.

PRE4. The prerequisite work is officially handed over
fiom prior trade.

PRES. The weather forecast is favorable for performing
work.

PRE6. Access to the work area (permits, etc) has been
obtained and cleared with respective party.

PRE7. All coordination issues with following trade have
been addressed by the project team.

PRE8. Overall, the prerequisite work is complete.

DIR1. The D&E documents for current work are
acceptable.

DIR2. The D&E documents for current work are
available.

DIR3. All RF I's for current work are addressed
satisfactorily.

DIR4. The owner is agreeable to any design/scope
changes, if any.

DIRS. The standard of work performance is available.

DIR6. The standard of work performance is clearly

-254-

Source
Liu and Ballard (2008)

Liu and Ballard (2008)

Author

Liu and Ballard (2008)

Liu and Ballard (2008)

Author

Dai. et. al. (2009)

Author

Dai eta]. (2009)

Dai et. al. (2009)

Dai et. al. (2009)

Author

Author

Author

expressed.
DIR7. The submittals are approved.
DIR8. The submittals are available.

DIR9. Final design instructions have been confirmed by
architect/engineer to avoid change.

DIR10. All lien waivers are in order.

DIR11. All permits are available.

DIR12. All lien waivers are submitted.

DIR13. Overall, all the directives are in order.
RES 1. The material is/will be delivered on time.

RES2. The material specifications match
contract/submittal specs.

RES3. Space is available for material lay-downs.
RES4. Path available for material transport.
RESS. The labor is informed of all work assignments.

RES6. The labor needed is available.

RES7. The tools to be used are in good working condition.

RESS. Right tools are available in enough quantity for
crews to work with.

RES9. The equipment to be used is in good working
condition.

RESIO. Right equipment is available in the desired
quantity.

RESI 1. Overall, all the required resources are available.

WTE1. The amount of work is sized appropriately to

Author
Author

Author

Author
Dai et. al. (2009)
Author
Author
Dai et. al. (2009)

Liu and Ballard (2008)

Author
Author
Dai et. al. (2009)
Author
Dai et. al. (2009)

Author

Dai et. al. (2009)

Author

Author

Horman and Thomas

avoid overproduction.

WTE2. Transportation of raw/processed materials
requires single handling (as opposed to more than once).

WTE3. Transportation of raw/processed materials is over
short distances.

WTE4. No amount of rework/correction is required for
completed work.

WTES. No over-processing (unnecessary finishing) is
performed for completed work.

WTE6. Inventory is delivered just-in-time to the
production unit.

WTE7. The production unit makes no unnecessary
movements to complete the work.

WTE8. The production unit is always working to
complete the work.

WTE9. The production crew is consulted on the best way
to perform the work.

WTE10. The production crew is consulted on the safest
way to perform the work.

WTE11. Overall, sufficient steps have been taken to
prevent generation of waste.

VAR1. There was no absenteeism in the workforce.

VAR2. The production crews operate with reserve
capacity.

VAR3. The learning curve effect is realizable for the
crews. ‘

VAR4. Recovery plans from occupational accidents are
standard practice.

-256-

(2005)

Ballard and Howell (1998)
Dai. Et. al. (2009)

Love et. a1. (1999)

Love et. al. (1999)
Thomas and Sanvido
(2000)

Author

Author

Dai et. al. (2009)

Author

Author

Dai. et. al. (2009)

Horman and Thomas
(2005)

Author

Rojas and Aramvareekul
(2003)

VARS. The production rates of the crew are consistent
over time (match Takt).

VAR6. The production plan has workable backlog in case
of interruptions to current assignment(s).

VAR7. The production plan has time buffers to prevent
interruptions to project completion.

BUR1. The work environment is free of potential
environmental hazards (air-home pathogens, dust,

chemical agents, ultra-violet light, and ionizing radiation).

BUR2. The work environment is comfortable to function
in (temperature and humidity, lighting, sun exposure,
noise levels). '

BUR3. The work environment is not congested.
BUR4. Overtime is not needed on our jobs.

BURS. The work is adequately paced to avoid physical
fatigue.

BUR6. The workers are equipped with PPEs, as
necessary.

BUR7. The workers are not allowed to engage work
involving heavy muscular loads.

BUR8. The workers have access to material-lifting
equipment.

BUR9. A Job Safety Analysis is performed.
BUR10. Workers are trained on proper lifting techniques.
BUR11. Rest cycles are built into the work method.

BUR12. Workers receive sufficient information regarding
the process flow and output.

BUR13. The rate of information does not exceed the

-257-

Author
Horman and Thomas
(2005)

Author

Author

Liu and Ballard (2008)

Dai et. al. (2009)
Author

Thomas and Raynar (1997)

Author

Author

Dai et. al. (2009)

Dai et. al. (2009)

Liu and Ballard (2008)

Thomas and Raynar (1997)

Dai et. al. (2009)

Thomas and Raynar (1997)

mental capacity of the worker to process.

BUR14. Identical or very similar signals don’t occur for a Author
long time.
BUR15. Adequate time is allowed for decisions and Author

resulting actions in the normal circumstances.

BUR16. Adequate time is allowed for decisions and Author
resulting actions in emergencies.

BUR17. Hand tools used are the correct ones for the task. Author
BUR18.Hand tools used are adequately maintained. Author
BUR19. All safety signs and visuals are correctly located. Dai et. al. (2009)
BUR20. Overall, the workers do not feel any kind of Author
burden.

-258-

 

REFERENCES

-259-

Arditi D. (1985). “Construction Productivity Improvement.” Journal of Construction
Division, Vol. 111(1), pp. 1—14.

Arditi, D., Tokdemir, OB. and Suh K. (2002). “Challenges in Line-of-Balance
Scheduling.” Journal of Construction Engineering and Management, ASCE, NY, Vol.
128 (6), pp. 545-556.

Bae BR. (2005). “LISREL structural equation model: Understanding & Application. ”
Chunglam Publishing Company, Korea.

Ballard G. (1997). “Lookahead Planning: The Missing Link in Production Control.”
Proceedings of the 5th Annual Conference of the International Group of Lean
Construction, IGLC-5, Gold Coast, Australia.

Ballard, G. (1999). “Improving Work Flow Reliability.” Proceedings of the 7’” Annual
Conference of the International Group of Lean Construction, IGLC-7, Berkeley, CA,
USA.

Ballard, G. (2000). “The Last Planner System of Production Control.” PhD. Thesis,
School of Civil Engineering, University of Birmingham, Edbagston, UK.

Ballard, G. and Howell, G. (1998). “Shielding Production: An Essential Step in
Production Control.” Journal of Construction Engineering and Management, ASCE, Vol.
124 (1), pp. 11-17.

Ballard, 0., and Tommelein, I., and Koskela, L., and Howell, G. (2002). "Lean
Construction Tools and Techniques." Section 15 in Rick Best and Gerard de Valence
(editors, 2002). Design and Construction: Building in Value. Butterworth-Heinemann,
Elsevier Science Ltd, pp. 227-255.

Barclay D., Higgins C., and Thomson R. (1995). “The Partial Least Squares Approach
(PLS) to Causal Modeling, Personal Computer Adoption and use as an Illustration.”
Technology Studies, Vol. 2(2), pp. 285-309.

Bentler P.M., and Chou C. (1987). “Practical Issues in Structural Modeling”.
Sociological Methods & Research, Sage Publications, Vol. 16 (1), pp. 78-117.

Bollen K., and Lennox R. (1991). “Conventional Wisdom on Measurement: A Structural
Equation Perspective.” Psychological Bulletin, Vol. 110(2), pp. 305-314.

Bollen KL. (1989). “Structural Equations with Latent Variables. ” John Wiley, New
York, USA.

-260-

Borcherding J ., and Oglesby C. (1974). “Construction Productivity and Job Satisfaction.”
Journal of Construction Division, Vol. 100(3), pp. 413-431.

Byme BM. (2006). “Structural Equation Modeling with E QS — Basic concepts,
Applications, and Programming (2nd edition)” Lawrence Erlbaum Associates, Mahwah,
New Jersey, USA.

Chua D.K.H., Kog Y.C., Loh PK, and Jaselkis EJ. (1997). “Model for Construction
Budget Performance — Neural Network Approach”. Journal of Construction Engineering
and Management, Vol. 123(3), pp. 214-222.

Coltman T., Devinney T.M., Midgley D.F., and Venaik S. (2008). “Formative versus
reflective measurement models: Two applications of formative measurement”. Journal of
Business Research, ScienceDirect, Vol. 61, pp. 1250-1262.

Dai J ., Goodrum P.M., and Maloney W.F. (2009b). “Construction Craft Workers’
Perceptions of the Factors Affecting Their'Productivity.” Journal of Construction
Engineering and Management, Vol. 135 (3), pp. 217-226.

Dai J., Goodrum P.M., Maloney W.F., and Srinivasan C. (2009a). “Latent Structures of
the Factors Affecting Construction Labor Productivity.” Journal of Construction
Engineering and Management, Vol. 135(5), pp. 397-406.

Davis F.D., Bagozzi RP, and Warshaw PR. (1989). “User Acceptance of Computer
Technology: A Comparison of Two Theoretical Models.” Management Science, Vol.
35(8), pp. 982-1003.

Diamantopoulos A. (2006). “The error term in formative measurement models:
interpretation and modeling implications”. Journal of Modeling in Measurement,
Emerald Group Publishing Limited, Vol. 1(1), pp. 7-17.

Diamantopoulos A., and Winkholfer HM. (2001). “Index Construction with Formative
Indicators: An Alternative to Scale Development.” Journal of Marketing Research, Vol.
38(2), pp. 269-277.

Diamantopoulos A., Riefler P., and Roth K.P. (2008). “Advancing Formative
Measurement Models”. Journal of Business Research, ScienceDirect, Vol. 61, pp. 1203-
1218.

DiekmannJ., and Heinz J. (2001). “Determinants of Jobsite Productivity.” Construction
Industry Institute Research Report No. 143-11, University of Texas, Austin.

Dvir D., Raz T., and Shenhar A.J. (2002). “An Empirical Analysis of the Relationship
between Project Planning and Project Success.” International Journal of Project
Management, Vol. 21(2), pp. 89-95.

-261-

Edwards J .R. (2001). “Multidimensional constructs in organizational behavior research:
an integrative analytical framework.” Organizational Research Methods, Sage

Publications, Vol. 4(2), pp. 144-192.

Edwards IR, and Bagozzi RP. (2000). “On the Nature and Direction of Relationships
between Constructs and Measures.” Psychological Methods, Vol. 5(2), pp. 155-174.

Faniran 0.0., Oluwoye J.O., and Lenard D. (1994). “Effective Construction Planning”.
Construction Management and Economics, Vol. 12, pp. 485-499.

F aniran 0.0., Oluwoye J.O., and Lenard D. (1998). “Interactions between Construction
Project Planning and Influence Factors.” Journal of Construction Engineering and
Management, Vol. 124(4), pp. 245-256.

Foulds, LR, and Quaddus, M., and West, M. (2007). “Structural Equation Modeling of
Large-scale Information System Application Development Productivity: The Hong Kong
Experience.” 6’“ IEEE International Conference on Computer and Information Science
(ICIS).

Hair J.F., Black W.C., Babin B.J., Anderson RE, and Tatham R.L. (2005). “Multivariate
Data Analysis (6“ Edition)”. Pearson Prentice Hall, Mahwah New Jersey, USA.

Hancock, 0. R., and Freeman, M. J. (2001). “Power and sample size for the root mean
square error of approximation test of not close fit in structural equation modeling.”
Educational and Psychological Measurement, Vol. 61(5), pp. 741-758.

Herbsman Z., and Ellis R. (1990). “Research of Factors influencing Construction
Productivity.” Construction Management and Economics, Vol. 8(1), pp. 49-61.

Horman M.J., and Thomas HR. (2005). “Role of Inventory Buffers in Construction
Labor Performance.” Journal of Construction Engineering and Management, Vol.
131(7), pp. 834-843.

Islam M.D.M., and F aniran O. O. (2005). “Structural Equation Model of Project Planning
Effectiveness”. Construction Management and Economics, Vol. 23, pp. 215-223.

Jarvis C.B., Mackenzie SB, and Podsakoff P.M.(2003). “A Critical Review of Construct
Indicators and Measurement Model Misspecification in Marketing and Consumer
Research.” Journal of Consumer Research, Vol. 30, pp. 199-218.

Kankainen, J. and Seppanen, O. (2003). “A Line-of-Balance Based Schedule Planning
and Control System.” Proceedings of the 11m Annual Conference of the International
Group of Lean Construction, IGLC-l l, Blacksburg, Virginia, USA.

Kim D.Y., Han SH, Kim H., and Park H. (2009). “Structuring the prediction model of
project performance for international construction projects: A comparative analysis”.
Expert Systems with Applications, ScienceDirect, Vol. 36, pp. 1961-1971.

Kim, Y.W. and Ballard, G. (2000). “Is the Eamed-Value Method an Enemy of Work
Flow?” Proceedings of the 8’“ Annual Conference of the International Group of Lean
Construction, IGLC-8, Brighton, UK.

Kim, Y.W. and Jang, J.W. (2006). “Applying Organizational Hierarchical Constraint
Analysis to Production Planning.” Proceedings of the 14’“ Annual Conference of the
International Group of Lean Construction, IGLC-l4, Santiago, Chile.

Kline, R. B. (2005). “Principles and practice of structural equation modeling (2nd
edition)” Guilford Press, New York, USA.

Liberda M., Ruwanpura J., and Jergeas G. (2003). “Construction Productivity
Improvement: A Study of Human, Management and External Issues.” Construction
Research Congress 2003, ASCE, Reston, VA.

Liker J.K. (2004). “The Toyota Way — 14 Management Principles from the World ’s
Greatest Manufacturer McGraw-Hill, USA.
Liu M., and Ballard G. (2008). “Improving labor productivity through production

control.” Proceedings of the 16’“ Annual Conference of the International Group of Lean
Construction, lGLC-l6, Manchester, UK.

Love P.E.D., Mandal P., and Li H. (1999).”Determining the causal structure of rework

influences in construction.” Construction Management and Economics, Vol. 17(4), pp.
505-517.

MacCallum RC, and Browne M.W. (1993). “The Use of Causal Indicators in
Covariance Structure Models: Some Practical Issues.” Psychological Bulletin, Vol.
114(3), pp. 533-541.

MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). “Power analysis and
determination of sample size for covariance structure modeling.” Psychological Methods,
Vol. 1(2), pp. 130-149.

MacCallum, RC, and Austin, IT. (2000). “Applications of Structural Equation
Modeling in Psychological Research.” Annu. Rev. Psychol, Vol. 51, Pg 201-226.

Downloaded from www.arioumals.annualreviews.org on 04/06/08.

MacKenzie S.B., Podaskoff P.M., and Jarvis CB. (2005). “The Problem of Measurement
Model Misspecification in Behavioral and Organizational Research and Some

Recommended Solutions.” Journal of Applied Psychology, Vol. 90 (4), pp. 710-730.

-263-

Maloney W.F. (1981). “Motivation in Construction: A Review.” Journal of Construction
Division, V01. 101(4), pp. 641-647.

Mitropoulos, RT. (2005). “Planned Work Ready: A Proactive Metric for Project
Control.” Proceedings of the 13’“ Annual Conference of the International Group of Lean
Construction, IGLC-13, Sydney, Australia.

Molenaar K., Washington S., and Diekmann, J. (2000). “Structural Equation Model of
Construction Contract Dispute Potential.” Journal of Construction Engineering and
Management, Vol. 126(4), pp. 268-277.

Oglesby 0, Parker H., and Howell G. (1989). “Productivity Improvement in
Construction.” McGraw-Hill, NY.

Purvis R.L., Sambamurthy V., and Zmud R.W. (2001). “The Assimilation of Knowledge
Platforms in Organizations: An Empirical Investigation.” Organization Science, Vol.
12(2), pp. 117-135.

Rojas M., and Aramvareekul P. (2003). “Labor Productivity Drivers and Opportunities in
the Construction Industry.” Journal of Construction Engineering and Management, Vol.

129(2), pp. 78-82.

Roberts N., and Thatcher J .B. (2009). “Conceptualizing and Testing Formative
Constructs: Tutorial and Annotated Example”. The DA TA BASE for Advances in
Information Systems, Vol. 40 (3), pp. 9-39.

Raykov, T., and Marcoulides, G. A. (2006). “A first course in structural equation
modeling (2nd edition)” Lawrence Erlbaum Associates, Mahwah, NJ, USA.

Ruiz D.M., Gremler D.D., Washburn J.H., and Carrion 0.0 (2008). “Service value
revisited: Specifying a higher-order, formative measure”. Journal of Business Research,
ScienceDirect, V01. 61, pp. 1278-1291.

Salem 0., and Solomon 1., and Genaidy A., and Luegring M. (2005). “Site
Implementation and Assessment of Lean Construction Techniques.” Lean Construction

Journal, Vol. 2(2), pg 1-20, USA.

Schumacker, R. E., & Lomax, R. G. (2004). “A Beginner's Guide to Structural Equation
Modeling. ” Mahwah: Lawrence Erlbaum Associates, USA.

Shehnar A. J ., and Laufer A. (1995). “Integrating Product and Project Management — A
New Synergistic Approach.” Engineering Management Journal, Vol. 7(3), pp. 11-15.

Singh, Y. (2007). “A Framework for Production Management of Renovation Projects.”
Master ’s Thesis, SPDC, Michigan State University, East Lansing, USA.

-264-

Thomas H.R., Horman M.J., Minchin Jr. RE, and Chen D. (2003). “Improving Labor
Flow Reliability for Better Productivity as Lean Construction Principle”. Journal of
Construction Engineering and Management, Vol. 129(3), pp. 251-261.

Thomas H. R., and Raynar K. A. (1997). “Scheduled Overtime and Labor Productivity: A
Quantitative Analysis.” Journal of Construction Engineering and Management, Vol.
123(2), pp. 181-188.

Thomas H. R., and Sakarcan A. (1994). “Forecasting Labor Productivity using Factor
Model.” Journal of Construction Engineering and Management, Vol. 120(1), pp. 228-
239.

Thomas H. R., and Sanvido V. (2000). “Role of Fabricator in Labor Productivity.”
Journal of Construction Engineering and Management, Vol. 126(5), pp. 358-365.

Ullman J.B. (2006). “Structural Equation Modeling: Reviewing the Basics and Moving
Forward.” Journal of Personality Assessment, Vol. 87(1), pp. 35-50.

United Nations Committee on Housing, Building, and Planning (United Nations) (1965).
Efi'ect of Repetition on Building Operations and Processes on Site, United Nations, NY.

Walt, L. (2006). “Earned Schedule Leads to Improved Forecasting.” Proceedings of the
PROA/LAC Conference, Sydney, Australia.

Wilcox J.B., Howell RD, and Breivik E. (2008). “Questions about formative
measurement.” Journal of Business Research, ScienceDirect, Vol. 61, pp. 1219-1228.

-265-