33 3333 33

3 ‘3 ' 3‘33.“ 333““ 33“, WW?

33“. - 3‘; '|‘$

i... .3 .égfl ’. ‘ h . I‘UI.
33.3.3.- 3. 33 3-
..(333.I t 4

I
. ”13"“

   
     
      

333“ h33. W3 (“'W 33333 3Q; g?
3 3' 333“ “:33 333mg,333133.3333331333‘3333333fi 333%,};
33333“ 33333333

3333.

  

 
 

  
 
 

33

3333'

33 333.333. {“333 33‘ 3’“ 3333333
333333

'33" 3333333“;
3'33 3"33333 3" ““33 3. “"33 33333 33333 333’
“3332;133 3 315:}

  

3 3

134!!3':“4‘ ’ i‘P

33- 3333 333
3333

   
 

  

   
   
     
    
 
        

3“

 

    

3 L 3: 3 39”}?! “3
. ' £4 ‘3 1' .
I 333333” 3 l: 33“.? 3E
3 '3 3'- 3
3 3 33:333.-
3 333
3333333
3 3 3333 .3

‘13:}
343131

  

  
 
 
  
   

  
  
 
  
 

 

  
 
   
   
       
  

 
      
 
 
 
    

 
 

  
   
 
 

 
 

.3 3 f
", 3.33 3. , 2 ”“11
3 ‘3 333 3 3 -' 3 13"
3 3.3“ h- 3
’33 “3” y 3 , 3“
3| .5 ;
3! )v’fi . “ ‘ ' l ‘1 ”327.1”: "
» , 3: ‘. , ‘3, 3. 333’s:
3:. 333‘" . 3 93333:
3 3 3.. 333 33333 .
Q 3; . . 311- 3.{; ti 3 ‘ - :3!
: i ‘ ‘ .~3333335333333335‘333~ . . ~3-
-- ' .“ .' 93M3333333- . ‘-- 3. .3
c , anyw3fi w: '33'V$fim-fifl 3 I
.. . " 3, ‘ 3333:3333 "'93 3 "3333 "33-3"; ‘-.
: 3 “3333 31“"; 3 . 2.3333 33.33.3333 3
a , 3 K3 .3J“g= v 3““3” H3. ,;
. ,3 3 3 3:33;. . '-'33-3 - 333? 3: 43‘: - 3"3
’3 - .9. .' ‘ 1.123331“ 3 3 3.3-. 33,13]:- 3‘:. "H "3"
3 ‘ 33 533335 33333333. ..,3:.:=:33.~,_3§3=333333;3.
.; 33 33. 1-333 ‘ 3'3 '

3 .33-333
, 33‘33‘1
13‘

3‘13”! w"

    
   
   
     
 

Ltd—M ~ 0 "J -.
fun-p:- "2...“... 4.3-... ..
- -_ .
’ ..... .a'.-’ -n M —n;

.

.. Wruz.
aé-S'gju‘gfi
w. -.

I .I.." ' '
.. ..
, . . .-.—
, .
.
.
.
.
__..;.___

  
  
 
 
 
    
    

pan-cu
‘ .‘H —»
erl‘“

  

 
  
    

   

£13 3.1331333“; 3

, 3,, 333.3. 3.
“.3’3“ .- 3h" .33! 3331(393'33311‘3 ’3
3, Mfim.3 3? fl?

3333331

3... 3: 3335‘.” ,;

3333
3333;3E3i1t:
-3 33

       
   
 
    
 
    

g \

  

3,:
.33 3

’3“W
33.33.

““33“ 3.333

3

 
 

3
““3“““33NM

3133 13.133333131333333
33333333 3333,1331.

33333 3 “333333

,- 3.3“ _ 33' ' 3:“
535-“.2'“ 3 '3 ”333‘ 33‘3‘3"

. .',"."."'.'3 "‘3 1““ 3|
.immJAMw33u ‘

 

3'4. t»..-_- 2:5

      
  

LIfiMBY
”.Erggigan ‘5th
L University

This is to certify that the

dissertation entitled

DEMOGRAPHIC AND FOOD RELATED DESCRIPTORS
OF DIET PROBLEM GROUPS IN THE
1977-78 NATIONWIDE FOOD CONSUMPTION SURVEY

presented by

AMY BAXTER SLONIM

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Human Nutrition

jor professor

7544’;an

Date July 24, 1982

 

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

 

MSU

LIBRARIES

 

 

 

DEMOGRAPHIC AND FOOD RELATED DESCRIPTORS
OF DIET PROBLEM GROUPS IN THE
l977-78 NATIONWIDE FOOD CONSUMPTION SURVEY

By

Amy Baxter Slonim

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

DOCTOR OF PHILOSOPHY

Department of Food Science and Human Nutrition

1982

ABSTRACT

DEMOGRAPHIC AND FOOD RELATED DESCRIPTORS OF DIET PROBLEM GROUPS
IN THE
1977-78 NATIONWIDE FOOD CONSUMPTION SURVEY

By Amy B. Slonim

Two questions in the 1977-78 Nationwide Food Consumption Survey
(NFCS) were about dieting behavior and factors affecting food intake.
About 50 percent of the 24,362 NFCS respondents reported at least one of
these dietary behaviors or factors.

Phase I of this study described respondents in terms of factors
affecting their food or nutrient intake. Respondents were categorized
into groups reporting medical and non-medical factors. These groups were:
NONE (50.5%), NON-MEDICAL (39.5%), NON-MEDICAL & MEDICAL (6.0%), and
MEDICAL (4.0%). Meal and snack patterns, demographic characteristics,
nutrient quality assessment, and food intake and related behaviors
comprised typologies for each group.

The groups with some type of medical problem, NON-MEDICAL & MEDICAL
and MEDICAL, were comparable in socio-economic descriptors and had the
most respondents 55 years of age and older. As expected, they lived in
smaller households with more: non-working adults, female only headed
households, and lower education levels for head of households. These
groups also were more similar in dietary intake from specific food groups

and other related behaviors (eating out and eating alone).

Amy Baxter Slonim

The NONE and NON-MEDICAL groups contained the most respondents less
than 18 years of age living in larger households with employed head(s) of
household. The personal food behaviors such as intake from specific food
groups, eating alone and eating out were more alike for these two groups
than the groups who identified a medical problem affecting intake.

The two groups identifying some type of non-medical factor affecting
intake, NON-MEDICAL and NON-MEDICAL & MEDICAL, had more respondents
ingesting less than 60 percent of their Recommended Dietary Allowances
(RDA) for seven nutrients. Conversely, the NONE and the MEDICAL groups
had more respondents ingesting nutrients at more than 59.9 percent of the
RDA. In addition, meal and snack patterns were more alike for NONE and
MEDICAL groups and for the groups identifying some type of non-medical
problem. It was concluded that meal and snack patterns and total mentions
of specific food groups were characteristics which differentiated
nutrient quality assessment of the four groups.

Phase II was a theoretical treatment of the data set. A model
representative of variable sets of factors potentially affecting food
intake was derived and estimated using multivariate techniques. The
independent variable sets represented demographic characteristics, food
related behaviors, and nutrient intake. Descriptive statistics were
generated between the sets of independent variables and the dependent
variable, problem versus no problem with dietary intake. In addition, the
relationships between the indicator variables representing the
independent variable sets were quantified.

The model was estimated using factor analyses, discriminant
analyses, and canonical correlation analyses. Low correlation
coefficients (Rc:§.3) were determined between the dependent variable and

each of the independent variable sets. Moderately high correlations

Amy Baxter Slonim

(Rc_>_J) were obtained between the sets of independent variables. The
Phase II analyses were critical steps in furthering derivation of
mathematical conceptual schemes to represent food related behaviors. The
findings from Phase II may be used to further refine and direct future

analyses to identify’ measured indicators of factors affecting food

intake.

To Louise Rose Wiener Slonim
and

my father and sister.

ii

ACKNOWLEDGMENTS

This dissertation has been contributed to by many persons in many
states. An attempt to recognize individuals who have influenced the
thought processes and/or analyses as well as my general mental health will
be organized by geographic locale. ‘

Starting in Nfichigan, Dr. Kathryn M. Kolasa 'planted the seeds',
guided my development and nurtured me through the past five years way
beyond the call of duty. I would like to thank her for fostering my
technical and philosophical development. Also in East Lansing, Dr. James
Bonnen and Dr. Carolyn Lackey, whose advise and guidance allowed me to
complete this dissertation. The MSU Department of Food Science and Human
Nutrition, the College of Human Ecology, and the Agricultural Experiment
Station have provided financial support through grants, fellowships, and
research assistantships. In addition, many members of the graduate
(community) nutrition group through the years have been exceptionally
supportive and solid sounding boards. In particular, Patricia Lynch, Dr.
Karen Penner, Bethann Hitcher, John Kallas, Jaci Fitzgerald and Mary Burke
have been there through the thick and thin. Barb Pumfrey Taylor whose
energy and sense of responsibility made the final days and final copies.

In Illinois and Missouri, past and present members of my guidance
committee, Dr. Kristen N. McNutt and Dr. Karen Morgan who have encouraged
and assisted me throughout.

Moving East to New York, Dr. Gilbert Leveille has challenged my
thinking to see the bigger picture and aided in the integration of

iii

nutritional science into my perspective. And in the heart of the 'Big
Apple', Dr. AnneMarie F. Crocetti (AFC), without whom this dissertation
would not have been. Much gratitude for sharing her wisdom, critical
thought processes, her data base, and her livingspace around the clock.
In addition, Francine Perlman and Carol Richmond in recognition of their
computer programming expertise and assistance and patience.

In Maryland, Dr. Edward L. Fink whose statistical consulting and
direction, kindness, and enthusiasm have been an inspiration behind the
completion of this project. And the rest of the Fink family who have
patiently shared their father and provided a home in Maryland. The
University of Maryland Computer Science Center and Mitchell Karpman in
particular for their continual computer technical knowledge support and
general concern.

In Washington, D.C., Dr. Catherine Noteki, Project Officer, USDA,
Consumer Nutrition Center for providing funding for this investigation.
Dr. Luise Light for connecting me with AFC and consulting along the way.

In D.C. and Virginia many friends and families have acted as
consultants and listened endlessly to aches and pains. In the last year
and a half, Kathy and Ralph Dawn and Margo Quiriconi in particular have
provided homes in my times of need. Thomas Thompson has been a diligent
editor and a primary support in these last months. And finally and
mostly, my Mom, Dad, Anne, Hunt, Jeffrey, and Randy whose unconditional

love and assistance made this a reality.

iv

TABLE OF CONTENTS

Page

LIST OF TABLES ..................... viii

LIST OF FIGURES ..................... xii

Chapter

I. INTRODUCTION ................... 1

Phase I .................... 2

Phase II .................... 3

II. REVIEW OF LITERATURE ............... 5

Nutrition Education Research in the l980's. . . 6

Use of Food Consumption Survey Data ...... 9
Multivariate Approaches to Understanding Food

Related Behaviors .............. 11

Summary .................... 13

Objectives of Investigation .......... 13

Phase I ................... 13

Phase II ................... 15

III. METHODS ..................... 16

Consequences of the Survey Design and Fieldwork 17

NFCS Sample Design .............. 18

Dietary Intake Methodology .......... 19

Wording of NFCS Questions .......... 21

Completion Rates of NFCS Respondents ..... 23

Summary ................... 23

Seasonality ................. 24

Phase 1: Methods ................ 24

Socio-Economic Characteristics ........ 25

Meal and Snack Patterns ........... 26

Nutrient Quality Assessment ......... 27

Food Group Intake and Personal Behaviors. . . 3o

Chapter Page

Analysis Phase I ............. 31
Phase II: Methods .............. 31
Description and Transformation of
Variables ................ 32
Tests for Multicollinearity ........ 35
Discriminant Analyses ........... 37
Canonical Analyses ............ 38
FOOTNOTES .................. 40
IV. RESULTS AND DISCUSSION: PHASE I ......... 41
The Sample .................. 41
Factor Intake Categories ........... 44
NON-MEDICAL Category ............ 46
NON-MEDICAL & MEDICAL Category ....... 48
MEDICAL Category .............. 53
Discussion: Intake Factor Category Groups. . 53

Identification of Meal and Snack Patterns,
Socio-Economic, Nutrient and Food

Characteristics of Factor Groupings ..... 55
Meal and Snack Patterns ........... 56
Socio-Economic Descriptors ......... 58
Nutrient Quality Assessment ......... 68
Nutrition Education Applications ...... 75
Food Group Intake and Personal Behaviors . . 77
Composite Typologies of Intake Factor
Category Groups .............. 85
Nutrition Education Implications ...... 87
V. RESULTS AND DISCUSSION: PHASE II ........ 91
Sample and Variable Descriptions ....... 92
Demographic Descriptors ........... 100
Nutrient Intake ............... 101
Food Related Behaviors ........... 103
Tests for Multicollinearity .......... 104
Discriminant Analyses ............. 110
Canonical Analyses .............. 113
Summary .................... 123
FOOTNOTE ................... 125

vi

Chapter

VI. SUMMARY AND CONCLUSIONS .............

Limitations of the Study ...........
Strengths of the Study ............
Conclusions/Implications ...........

APPENDICES
APPENDIX

Phase I ...................
Phase II ..................

A. Variable Construction and Descriptions .....

1.

boom

5.

Calculations for Development of
Marginality Score ...........
Calculations for Development of PFC Score
Crocetti 32 Food Groups .........
Glossary of Independent Variable Set
Indicators for Phase II ........
Description of Composite Index ......

B. Principal Component Factor Analyses and
Correlation Tables ..............

REFERENCES . . .

vii

Page
126
130
131
131

131
133

134

134
135
136
137
142

143

153

LIST OF TABLES

Table
1. Sequential Deletion of NFCS Individuals .......

2. NON-MEDICAL Group by Number and Type of Reported
Factor(s) .....................

3. NON-MEDICAL group by Sex/Race and Number of Reported
Factor(s) .....................

4. NON-MEDICAL 8 MEDICAL Group by Number and Type of
Reported Non-Medical Factor(s) ..........

5. NON-MEDICAL & MEDICAL Group by Total Number of
Non-Medical Factor(s) ...............

6. NON-MEDICAL 8 MEDICAL Group by Number and Type of
Medical Problem ..................

7. NON-MEDICAL & MEDICAL Group by Sex/Race and Type of
Non-Medical Factors ................

8. MEDICAL Group by Number and Type of Medical

Problem(s) ....................
9. Intake Factor Groups by Meal Patterns ........
10. Intake Factor Groups by Snacking Behavior ......

11. Intake Factor Groups by Census Age Categories. . .

12. Intake Factor Groups by Size of Household ......
13. Intake Factor Groups by Family Composition .....
14. Intake Factor Groups by Relationship of Respondent
to Head of Household ...............
15. Intake Factor Groups by Poverty Level Index .....

16. Intake Factor Groups by Education of Head of
Household .....................

viii

Page
42

47

49

50

51

51

52

54
57
57
59
60
6O

61
62

63

Table
17.

18.
19.
20.
21.
22.
23.

24.

25.

26.

27.
28.

29.
30.
31.

32.

33.
34.
35.

36.

Intake Factor Groups by Working Status of Head(s)
of Household ...................

Intake Factor Groups by Four Census Regions .....
Intake Factor Groups by Marginality Index ......
Intake Factor Groups by PFC Index .........
Intake Factor Groups by Composite Index .......
Sex and Race by Composite Index ...........

Snacking Behavior and Four Major Meal Patterns by
Six Representative Composite Index Categories. . .

Education Levels of Head of Household by Composite
Index .......................

Crocetti Food Groups by Total Number of Mentions for
Three Days ....................

Intake Factor Groups by Total Number of Mentions for
Fluid Milk for Three Days .............

Intake Factor Category Groups by Eating Out Behavior

Intake Factor Category Groups by Eating Alone
Behavior .....................

Intake Factor Category Groups by Supplement Use. . .
Intake Factor Category Groups by Weight Status . . .

Descriptive Statistics for Indicators of Independent
Variable Sets ...................

Demographic Characteristic Indicators by Problem
Group .......................

Nutrient Quality Indicators by Problem Groups. . . .
Food Related Behavior Indicators by Problem Groups .

Highly Correlated Pairs of Nutrient Quality
Indicators ....................

Tests for Equality for Dependent Group Covariance
Matrices for Independent Variable Sets ......

ix

Page

64
65
69
69
72
72

73

74

78

79
81

82
82
83

93

97
98
99

106

112

Table
37.

38.

39.

40.

41.

42.

43.

8-1.

8-2.

8-3.

8-4.

8-5.

8-6.

8-7.

B-8.

Canonical Discriminant Functions for Problem Group
by Each Independent Variable Set ........

Canonical Correlations Among Indicators of
Demographic Characteristics and Nutrient Quality

Canonical Correlations Among Indicators of
Demographic Characteristics and Food Related
Behaviors ...................

Canonical Correlations Among Indicators of
Nutrient Quality and Food Related Behaviors . .

Canonical Variate Loadings for Indicators of
Demographic Characteristics and Nutrient
Quality ....................

Canonical Variate Loadings for Indicators of
Demographic Characteristics and Food Related
Behaviors ...................

Canonical Variate Loadings for Indicators of
Nutrient Quality and Food Related Behaviors . .

Correlation Matrix for Indicators of Demographic
Characteristics ................

Correlation Matrix for Indicators of Nutrient
Quality ....................

Correlation Matrix for Indicators of Food Related
Behaviors ...................

Rotated Factor Structure of Indicators of
Demographic Characteristics ..........

Rotated Factor Structure of Indicators of
Nutrient Quality ................

Rotated Factor Structure of Indicators of
Food Related Behaviors .............

Correlations Among Indicators of Demographic
Characteristics and Nutrient Quality ......

Correlations Among Indicators of Demographic
Characteristics and Food Related Behaviors. . .

Page

112

114

115

116

119

120

121

143

144

145

146

147

148

149

150

Table Page

B-9. Correlations Among the Indicators of Nutrient
Quality and Food Related Behaviors ..... 151

xi

LIST OF FIGURES

Figure
1. Model to be Quantified in Phase II .........
2. Intake Factor Category Groups in Phase I ......

3. Descriptive Composites of Intake Factor Groups . . .

4. Model Including Indicators of Independent Variable
Sets .......................

5. Estimated Model Based on First Canonical Variates. .

xii

Page

14
45
86

109
118

CHAPTER I
INTRODUCTION

In the last decade, the nutritional adequacy of the American diet has
received attention in the mass media. Women's and men's magazines not
only include meal planning and preparation articles, but also Inore
technical articles on food, nutrition and dieting. Newspaper food editors
also cover relevant nutrition issues and book stores stock food and
nutrition books.

To begin to assess the U.S. general population's concern about
personal food and nutrient intake patterns or problems, the U.S.
Department of Agriculture (USDA) staff included two new questions in the
1977-78 Nationwide Food Consumption Survey (NFCS) (Cronin, 1980). The
first question gave individuals an opportunity to report whether or not
they were on a doctor prescribed special diet, on a group, or on an
individual diet regimen. The second question identified nine items that
might affect what a person eats. or' drinks and gave respondents an

opportunity to check as many as pertained. These nine items included:

I'm on a diet to lose weight

- I'm on a diet to put on weight

I have a chewing problem because of teeth

I have a medical problem like diabetes or allergy

Some foods do not agree with me

I don't feel like eating breakfast early in the
morning

- I have no interest in cooking for one person

- I do not like certain foods

Other

Crocetti and Guthrie (1982) conducted a secondary analysis of the
NFCS to explore changes in lifestyle and associated characteristics of the
diet and nutrient adequacy of respondents. They found approximately 50
percent of the respondents in the Spring quarter of the survey falling
into one or more of the above categories. The large percentage of
respondents who placed themselves in these categories afforded a unique
opportunity to begin to identify and characterize persons with medical
and/or non-medical practices or problems that they perceived as affecting
the way they ate or drank. A primary objective of this investigation was
to identify and characterize persons who self-reported medical and non-
medical factors 'hi the NFCS. The large sample size (approximately
25,000), the collection of data over an entire calendar year (four
quarters), and the combination of data obtained on demographics, nutrient
intake and food related behaviors added to the uniqueness of this
investigation. A second objective of this study was to use statistical
methodologies to derive a model incorporating four sets of variables: (I)
identification of problem affecting food intake; (2) demographics of
respondents; (3) personal and food related behaviors; and (4) nutrient
intake. Multivariate analysis techniques were used in: explore the
correlational relationships between these sets of variables
characterizing food related behaviors. The analyses occurred in two

phases.

Phase I
In Phase I after the sample for the anlyses was determined, the
respondents were grouped into four categories based on reported factors
which may have affected their food consumption. The four factor intake

categories were: NON-MEDICAL; NON-MEDICAL and NEDICAL; MEDICAL; and

NONE. Variables were used directly from USDA NFCS codes or were
constructed to describe the four groups in terms of: (1) socio-economic
characteristics; (2) meal and snack patterns; (3) nutrient quality
assessment; and (4) food group intake or related personal behaviors.
Traditionally the data from USDA surveys have been used to
characterize households and individuals by nutrients consumed by age,
sex, region, income, household size, or some combination of these
variables. Nutrition education efforts have been criticized for failure
to recognize changes in the nature and composition of the food supply and
failure to address target populations in relevant social, demographic and
lifestyle patterns. Phase I was designed to look for an alternative way
to analyze the NFCS data. It was thought that typologies of food
consumption patterns might be found among diet problem groups of
respondents. These typologies were constructed to provide useful

information to nutrition educators.

Phase 11

Phase II was a theoretical treatment of the NFCS data set. The
relationships between sets of independent variables and the dependent
variable, identification of factor or problem with dietary intake were
estimated. The independent variable sets represented demographic
characteristics, food related behaviors, and nutrient intake.
Mulitvariate analysis techniques were used to estimate the mathematical
relationships between the four sets of variables. Two five percent random
subsamples of the total study population (24,362 respondents) were
investigated for purpose of cross-validation. The end result was a model
representing the correlational relationships between the sets of

independent variables and the dichotomous dependent variable problem

versus no problem with intake.

CHAPTER II
REVIEW OF LITERATURE

A variety of surveys: consumer expenditure; household and
individual food consumption; and nutritional status have been used to
describe food intake patterns of individuals and households. Regardless
of survey size, an underlying objective has been to assess dietary intake.
The emphases and uses have been as varied as appraising nutrient intake of
specific segments of the population, providing baseline data for
development of policies and programs on consumer education, nutrition,
and food and agriculture, and deriving marketing strategies and consumer
product development by food industries. The use of food consumption
survey data for practical and theoretical nutrition education research is
discussed in this chapter.

The 1940's Committee on Food Habits (NRC, 1945) encouraged some of
the first research in the area of food habits. In 1964, Mead noted little
progress in theories or methodologies for conducting food habit research.
She proposed a multi-dimensional code for describing dietary patterns in
physiological, sensory, chemical, nutrition and cultural terms. Almost
two decades later, minimal progress has been made in defining the
relationship between independent and dependent variables affecting food
choices and behavior. A state of the art regarding the development and
direction of methodologies applied to describe and quantify food related

behaviors is forthcoming in a report of the National Academy of Sciences

Panel FActors Affecting Food Consumption (Kolasa, Lackey and Slonim,
1981).

Food habit research has been conducted incorporating multivariate
approaches with varying degrees of success using the theoretical,
scientific, and practical expertise of nutritionists, anthropologists,
economists, psychologists, and sociologists. The usefulness of
multivariate techniques in discovering regularities in the behavior of
two or more variables are described in this chapter. Additionally, the
model incorporating demographic, nutrient quality, and food intake

variables estimated in Phase II of the analyses is presented.

Nutrition Education Research in the 1980's

As we move into the 1980's, nutrition educators are being challenged
to build on traditional methods of research and information dissemination
with innovative and more effective techniques. The federal government has
fostered and supported this goal by sponsoring national conferences such
as the 1979 National Conference on Nutrition Education: Directions for
the 1980's (U.S. Department of Health, Education and Welfare et al.,
1980). The purpose of the Conference was to provide direction and
guidance in the form of recommendations, options and priorities to the
sponsoring groups and other public, private and voluntary agencies
addressing nutrition education needs for the 1980's.

The U.S. Department of Agriculture (USDA) has further demonstrated
its commitment by sponsoring a series of workshops at Pennsylvania State
University in 1980 to identify priority research issues in nutrition
education. The topics of the conferences were: Eating Patterns;
Nutrition Communication; Formal Nutrition Education; and Community

Nutrition Education. The goals of the Conferences included: (1) defining

and delimiting discrete areas of research encompassed in specific areas of
nutrition education research; (2) determining methodological and
conceptual problems currently limiting work in these areas; and (3)
identifying more fruitful directions for future research efforts (Sims,
1980).

In the specific recommendations of the task forces from the National
Conference held in September, 1979 (Dwyer, 1980) each group emphasized the
need to focus research to gather relevant information from specific
segments of the population on food habits, beliefs and related behaviors
to be able to target messages more appropriately. Nutrition messages,
regardless of their form, must be meaningful to the target groups within
their cultural, social, and economic orientations. It has been concluded
(Olson and Gillespie, 1981; Sanjur, 1982) that research methodologies and
data analyses need to be bolstered to gain insight into individual's or
group's food related behaviors.

Bass, Wakefield and Kolasa (1979) defined food behavior as an
individual's response to stimuli related to the selection, procurement,
distribution, manipulation, storage, consumption and disposal of food.
The food that people choose to eat, the reason for their choices, and
their eating patterns (frequency, eating partners, location) are
behaviors nutrition professionals have sought 11> understand. Many
studies have indicated that food and nutrient intake behavior is
associated with several interacting factors such as income, education
level, culture, socialization, geographic location, composition of family
and life cycle stage. The relationship between these factors and whether
or not a person is on a special diet (medically or otherwise prescribed)
or has some personal or non-medical factor (i.e., chewing problem or food

dislikes) which affects his/her food consumption behavior has not been

explored. Analyses of this nature may provide valuable information to
professionals in federal, state, or private agencies developing nutrition
education tools for consumers; the food industry interested in product
development and marketing; and/or legislators in determining and
administering programs and policies.

The Nationwide Food Consumption Survey (NFCS) for the first time
included two questions which gave individuals an opportunity to report
whether or not they were on a doctor prescribed special diet, on a group,
or self determined diet regime. USDA also included nine items which
allowed respondents to identify factors which may have affected the way
they ate or drank. These items included such factors as: being on a diet
to lose or gain weight; having a chewing problem; not liking to eat
certain foods or breakfast; and foods not agreeing with them. Respondents
were asked to check as many as applied to their intake.

To date analyses of USDA survey data have not specifically studied
persons on special diets or having self-reported factors affecting their
intake. In the highlights from a national workshop on nutrition education
research, (Olson and Gillespie, 1981) research priorities for the future
were enumerated. Among the prioritized areas for research were the
identification of lifestyle factors influencing food choice and dietary
behavior and factors in the affective domain influencing dietary
behavior. The NFCS afforded an opportunity to describe and quantify
demographic and food and nutrient intakes of persons who self-reported
problems with their intake that would be current and useful to nutrition

educators.

Use of Food Consumption Survey Data

The U.S. government has been responsible for measuring and
appraising trends in the U.S. food consumption since the 19th century.
Marr (1971) and Pao (1977) traced the development of dietary standards and
methodologies used to assess household and individual food consumption
from the 19th century European analyses to the 1977-78 USDA NFCS. The
data obtained from these investigations traditionally have been used in
part to identify the foods that people choose to eat and the subsequent
nutrient intake, eating patterns (frequency, eating partners, location),
and the relationship of foods/nutrients consumed with age, sex, race,
income and other demographic characteristics (Aquwa, 1980).

The results have been used by federal agencies, the food industry,
and research and educational institutions. Clark (1974) classified the
potential uses of data from nationwide food consumption surveys into four
categories: (1) appraisals of food consumption and dietary adequacy; (2)
control and regulatory uses (i.e. effects of enrichment of foods); (3)
food budgets and guidelines; and (4) economic, marketing and nutrition
research (which impact on the development and administration of
government programs and policies). Incorporation of individual's or
group's perceptions of non-medical factors (social or behavioral) or
medical problems that affected their food habits may further enhance the
potential uses of these surveys. For example, guidelines may specifically
be developed to include messages or terms relevant to population segment's
perceptions of problems or factors affecting their intake. Or
understanding and appraising dietary adequacy of the sample population
may be conducted in groupings based on identified medical or non-medical
problems. The findings may aid in identifying marketing strategies or

applied nutrition research relevant to people's perceptions of factors

10

affecting their intake.

Jenkins (1982) included an extensive review of dietary and food guide
development in the U.S. Historically, USDA has developed food selection
guidelines with the objective of translating dietary standards into
simple and reliable nutrition education tools useful to consumers in
satisfying their nutritional needs (Light and Cronin, 1981). The first
so-called food guides were attributed to Caroline L. Hunt who developed "A
Week's Food for the Average Family" published in 1921 by USDA and the 1923
bulletin entitled "Good Proportions in the Diet" (Hill and Cleveland,
1970). Since that time USDA has published several food selection guidance
tools utilizing the following data sources for updating: nutritional and
dietary status of the population, nutritional standards, food consumption
patterns, food availability, nutritive composition of foods and food
economics (USDA-Consumer and Food Economics Institute, 1976).

In 1976, the Consumer and Food Economics Institute held discussions
on the food selection tools developed to date. In review of commentaries
and critiques of the subject, the criticisms were summarized into three
broad subject categories (Light, 1977):

l. failure to address the most important. public
health nutrition problems

2. failure to recognize changes in the nature and
composition of the food supply

3. failure to recognize changes in social and
demographic characteristics and lifestyles of the
population

More recently a series of articles were published which discussed food
guidance for the public (Guthrie and Scheer, 1981; Dodds, 1981;
Pennington, 1981; Lachance, 1981; Light and Cronin, 1981). Varying

methods for developing guidance plans and specific suggested guidance

tools were presented. Each author emphasized incorporating current

11

consumption patterns and food acceptability to population segment in any
food guidance for the general public.

Phase I of this investigation was designed to describe social and
demographic characteristics of the study population. Furthermore,
nutrient quality assessment variables were constructed (Chapter III) to
better address relevant public health nutrition problems. The U.S.
Surgeon General's Report on Health Promotion and Disease Prevention
(1979) indicated decreasing incidence of nutritional deficiencies due to
insufficient intakes of vitamins and minerals. Current dietary concerns
in the U.S. have related to excessive intakes of certain macronutrients or
unbalanced intakes of macronutrients.

As our knowledge of nutrition has expanded it has
become more appropriate to emphasize, for dietary
guidance purposes, the energy producing nutrients,
protein, fat and carbohydrate, since excess of these

may be related to some of the more prevalent chronic
diseases in our society today (Jenkins, 1982:15).

Multivariate Approaches to Understanding Food Related Behaviors

The scientific study of human nutrition, like any other science, has
been fundamentally concerned with establishing laws of relationships
among factors given certain conditions (Monge, 1980). Nutrition science
has been concerned with the body's need for nutrients and how these
nutrients function in biochemical mechanisms. The application of
nutrition science in clarifying food related behaviors has necessitated
the incorporation of various environmental external factors and internal
factors in deriving conceptual frameworks or models. The formulation of
laws relating variables has been a theoretical endeavor dependent upon
empirical techniques. The application of mathematics to this process has

aided 'hi the: (1) identificathmw of consistent relationships among

12

variables; (2) understanding of complex information in a concise and
meaningful way; and (3) creation of derivations which are content free and
have allowed predictive capabilities which may be tested (Fink, 1979).

Multivariate analysis techniques have been used by many disciplines
in discovering regularities in behaviors of two or more variables. These
techniques have facilitated the development of 'multivariate profiles'
which have grounded understandings of relationships between variables for
model and theory development and testing. Multivariate analysis
techniques have been built from nethematical methods including matrix
algebra, geometry, the calculus, and statistics.

The consensus in the nutrition professional community has been that
more adequate theories are needed related to food behaviors (Olson and
Gillespie, 1981). Blalock (1969) noted that "theories do not consist
entirely of conceptual schemes or typologies, but contain lawlike
propositions that interrelate the concepts or variables two or more at a
time." A short run goal of theory development may include the process of
finding predictor variables causally related to the variable(s) to be
explained. However, in the long run it is theory that will provide the
terms by which complex interrelationships may be explained. As Woelfel
and Fink (1980) discussed, mathematics may be helpful in various stages of
theory building in understanding complex information in rich and
simplified ways. Relationships among variables may be derived from
mathematics 'hi content free terms which allow prediction and eventual
modeling and testing.

Phase II of this investigation was designed to quantify the
relationships between sets of independent variables representing
demographic characteristics, food related behaviors and nutrient intake

and the dependent variable, identification of problem with intake (See

13

Figure l). Multivariate analysis techniques were used to idiscover
'multivariate profiles' of regularities in behavior among variables. The
findings from this investigation may aid in further defining explanatory
variables and causal relationships between factors related to food
behaviors. It is through grounded conceptual schemes and deductive
reasoning that theories will be derived in the field of applied nutrition
science. The analyses in Phase II are a step in the direction of
grounding conceptual schemes, through the derived mathematic

representation of relationships between sets of variables.

Summary
Although the Committee on Food Habits in 1945 encouraged multi-

dimensional approaches to the study of food related behaviors, little
progress has been noted in the development of theories in this arena.
Multivariate analyses techniques have been used with varying degrees of
success in furthering theoretical gains in the applied field of nutrition.
The model investigated in this study incorporated understandings
previously derived between variables affecting food related behaviors
(Kolasa, Lackey and Slonim, 1981). Mathematical techniques were used to
quantify relationships among sets of variables to ground conceptual
schemes and specific factors which are interrelated in food habits.
Findings from this investigation may have nutrition education and theory

building implications.

Objectives of Investigation
Phase I
1. To identify NFCS respondents four years of age and older who self-

reported a medical or non-medical problem which may have affected

14

.HH omega cw umpmswpmm mg on choz .H mmstu

       
 
         

mco.><zmn
Dunk-.mp—
GOO".

mx<hy= It;
ENJOOKQ .0
ZO.P<O.u—_hzmn=

Oin<¢002mn

15

their eating pattern or nutrient intake.

2. To characterize respondents identified in #1 above in terms of socio-
economic factors, meal and snack patterns, nutrient intake
assessment, and food and related personal behaviors.

3. To discuss implications from #1 and #2 above as they relate to
nutrition educaton and research.

Phase II

4. To identify sets of indicator variables representing: demographics,
nutrient intake, and food related behaviors of NFCS respondents.

5. To quantify relationships between sets of independent variables
identified in #4 above and the dichotomous dependent variable,
identification of problem affecting intake.

6. To apply canonical analyses to determine relationships between the

three sets of independent variables identified in #4 above used to

characterize food behaviors.

CHAPTER III
METHODS

When conducting a secondary analysis of data originally collected
for other purposes there are issues that need examination. The data may
not be in a format best suited to the proposed secondary analyses. This
chapter includes discussion of several points about the design, field
work, coding and editing of the Nationwide Food Consumption Survey (NFCS)
which affected subsequent analyses. and ‘their interpretation. These
include the decisions to: not weight the individual sample to correct
representiveness; merge the records for three days of food intake into a
single record; exclude selected variables due to wording in questionnaire
or due to non-completion rates; and merge the data for the four quarter
subsamples into one data set.

Additionally, this chapter includes the methods used in Phase I to
select a sample and categorize respondents into ‘four intake factor
categories. The variables used to characterize the four groups in terms
of socio-economic, meal and snack patterns, nutrient quality assessment,
and food related behaviors are then discussed. The recoding and labeling
of meal occasions and multiple meals is described. The most frequent
patterns of number of meals eaten in each of three days was used to derive
the meal pattern variable. The nutrient quality assessment measurement
incorporated two indices. One identified the number of nutrients ingested

at less than or equal to 59.9 percent of the Recommended Dietary

16

17

Allowances (RDA) (NAS Food and Nutrition Board, 1980); this is called a
marginality score. The other considered total caloric intake of
respondents in proportions of protein, fats, and carbohydrates, a PFC
score. The combination of the marginality score and the PFC score yielded
a composite index. Variables constructed to describe food related
behaviors included: 32 food group categories; a combination of meal and
snack eating-out behavior; use of vitamin and mineral supplements; three
day behavior of eating alone; and weight status.

The final section of this chapter includes the sample and variable
selection and statistical analyses used in Phase II. The preliminary
examination and transformation of variables is discussed. Three types of
multivariate analysis techniques were used to estimate the model
incorporating four sets of variables: (I) identification of problem
affecting intake; (2) demographics of respondents; (3) personal and food
related behaviors; and (4) nutrient intake. The multivariate
methodologies used were discriminant analysis, factor analysis, and

canonical correlational analysis.

Consequences of the Survey Design and Fieldwork

In a Report to the Select Committee on Nutrition and Human Needs, the
U.S. General Accounting Office (1977) described the history, funding, and
eventual contracting of the NFCS and listed three consequences of the
design. As a secondary analysis researcher, these three consequences were
considered in the project design, analyses and interpretation. First, the
sampling scheme of households did not yield a representative sample of
individuals in the U.S. Secondly, the three day dietary intake technique
used to assess individuals food consumption was not validated. And,

finally, the wording and response options for several of the questions

18

limited the potential for the clarity and analysis of the data obtained.
Each of these consequences will be described separately and the

implications identified.

NFCS Sample Design

 

A random probability multistage, stratified sampling design intended

to represent all private households in the 48 adjacent states was used in

 

the Survey. The sample design imposed limitations on utilizing the
individual data for analysis. The design was based on the household as
the sampling unit; therefore, the probability of selection of a sampling
unit was limited to the household and did not apply to individuals within
households. Furthermore, the individual data were inspected to see if
they represented the 1977 U.S. population age and sex distribution as
determined by the U.S. Bureau of the Census (1978). ANAREM (1980)
concluded that the NFCS respondents did not yield a representative age/sex
sample of individuals in the United States in 1977. The selection of
individuals varied from interviewing wave to wave. For example, all
household members were to be interviewed in the spring quarter, but then
only half of the persons 19 years of age and older were to be interviewed
within each household in the other three quarters. The result of this was
that different proportions of children and adults in the populations were
found between the first and later three quarters.

Beyond the selection process, completion rates also affected the
representativeness of the sample. The non-compliance rate was not random.
Crocetti and Guthrie (1980) investigated the completion rates for the
individual sample. They found that young men and large households had

disportionately high non-compliance rates.

19

In addition, the recipients of contracts to conduct secondary
analyses did not receive a complete report on the sampling process
employed by fieldworkers. To ascertain the impact of the design and field
work execution on the individual sample, the following would be needed for
each stratum for each quarter: (1) sample size by number of members in
the household; (2) breakdown of reasons why questionnaires were not
completed; and (3) demographics on neighborhoods of completed and non-
completed individual questionnaires. This information has not been
supplied to date and is not expected.

Weighting is a means for correcting non-representativeness in sample
populations. USDA provided contractees with a set of weights recommended
for assignment to various strata of individual respondents. Since USDA
provided no explanation or formula for the derivation of the weights
recommended, they were not used in these analyses.

Due to the sample design and the limited information provided on the
field work execution and the non-compliance rates, the individuals
surveyed are not a random sample of the U.S. population in 1977. The NFCS
respondents are, however, a describable population. Therefore, the
typologies determined in Phase I were carefully identified in terms of
socio-economic status for each group. Characterizations were made based
on the findings of the analyses and not with regard to representativeness

of the U.S. population.

Dietary Intake Methodology

The second major design weakness that the Comptroller General U.S.
(General Accounting Office, 1977) stressed was the lack of validation in
the dietary intake methodology. In choosing methods for obtaining food

consumption data there are certain practical considerations which must be

20

addressed. These include such matters as number of respondents to be
surveyed and their required cooperation and time, the necessary training
for field workers, the resources available for coding and analyses, and
the objectives of the study (Caliendo, 1980).

The dietary intake methodology used in the individual questionnaires
included one 24-hour dietary recall and two-day food records on three
consecutive days. The 24-hour recall relied on the memory of the
respondent and a trained interviewer to obtain the data. The two-day
records required respondents to record every item of food or beverage
consumed on a special record sheet with indication of the amount ingested,
time consumed, and a label for the kind of meal or snack intake.

The GAO Report (1977) indicated that nine different methods for
collecting information from households and individuals were investigated.
The eventual techniques were selected since they yielded data "in the mid-
range of the array of results" (U.S. General Accounting Office, 1977).
The selection process used does not suggest anything about the validity of
the different methods or how the methods differ from one another. The
concluding comment of the Report related to dietary intake methodology'was
that in general procedures should be validated prior to usage in a survey.

Researchers have investigated the validity and sources of variance
in various dietary intake methodologies (Murray, 1970; Marr, 1971; Burk
and Pao, 1976; Houser and Bebb, 1978; Beaton et al., 1979; Stunkard and
Waxman, 1981; Karvatti and Knuts, 1981; and Carter, Sharbaugh and Stapell,
1981). For example, Beaton et al. (1979) discussed the sources of
variance in 24-hour recall data and found significant differences in the
day of the week interviewed. This would have been avoided if field
workers had obtained an equal distribution of the days of the week in

which initial interviews were conducted. However, upon inspection of the

21

data Crocetti and Guthrie (1980, 1982) found the weekends were
underrepresented. An even distribution across all days of the week would
result in approximately 14.3 percent (1/7th) representation per day of
week. In the Spring quarter, Saturday accounted for only 3.3 percent and
Friday 10 percent of the total day of the week distribution. The combined
Friday/Saturday (13.3%) was less than the desired 14.3 percent per day.
In addition, reports on the problems and outcomes of the 24-hour recall
and the two-day record suggested that systematic bias may be introduced by
the recall method. The 24-hour recall may produce fewer food items than
the food record method.

In light of the concerns discussed above, in the present study data
for the three day individuals food intake were merged and an average of
the attributes was used in all analyses. Pooling the three day data on
food intake offered advantages in smoothing out daily variations. The
large sample size and the merging of the data for the three days aided in

the reliable use of these data.

Wording of NFCS Questions

The third area of criticism from the GAO Report (1977) related to the
wording and response options of the actual questions in the
questionnaires. To begin with, several of the variables were ascribed to
households rather than individuals. For example, education, working
status, and occupation were obtained for only the head(s) of households.
In assessing demographics of the population, variables would have proved
more representative of actual respondents if they had been obtained for
more than head(s) of household. Generational differences may exist, but
it was not possible to assess them. This limited the interpretation of

the data.

22

In addition to being ascribed only to the head(s) of household, the
occupation variable was not categorized on the basis of the U.S. Labor

Dictionary of Occupational Titles (1977). The categories in the

 

Dictionary are established according to levels of skill and job
responsibility and according to industry, in order to reflect income,
lifestyle and responsibilities of persons in the categories. The NFCS
provided respondents seven options which mixed unskilled and skilled
labor categories. As a result, the variable for occupation did not
differentiate respondents well based on their jobs and was not used in
this investigation.

Several questions in the survey form for individuals had the category
"other" as an option. Included were such questions as: "Reason why the
day of intake was not normal"; "Race"; “Are you on a special diet? Iflygs,
how would you describe it?" and "These are some things that might affect
what a person eats and drinks. Indicate which ones, if any, pertain to
you." Crocetti and Guthrie (1980) found that in the Spring quarter, of
those respondents that reported a non-normal Day 1, 65 and 63 percents of
the black females and males, respectively, gave "other" as the reason.
Forty-two percent of the white males and females reported "other" as their
response. There was no way reliably to interpret such questions. If
these questions had been pretested effectively, the large percentages of
individuals in the "other" categories could have been avoided. Some
question responses had to be omitted from analyses in this study (for
example “Was this a normal intake day?") due to the large percentages of

respondents placing themselves in the "other" category option.

23

Completion Rates of NFCS Respondents

The final problems encountered in utilizing the NFCS data involved
low completion rates and under-reporting of intake items. In the Spring
Quarter, Crocetti and Guthrie (1981a.) found that only 65 percent of the
respondents completed the three questions on education, occupation and
income. In view of the low completion rates an attempt to establish a
socio-economic status index had to be abandoned.

There was no accurate way to estimate the level of under-reporting
for particular food items, except to compare these data with other
reports. For example, the U.S. Surgeon General's Report Healthy People
(1978) and Hyman et al. (1980) indicated substantially higher per capita
consumption of alcohol than was found in the initial analyses of the NFCS.
In addition, the percentages of respondents reporting candy consumption
appear low in relation to other national data. "It may have been due to a
reluctance to report an item that is popularly classified as a 'junk food'
or because candy may be popped into the mouth and forgotten" (Cronin,

Krebs-Smith, Wyse and Light, in press 1982).

Summary

In summary, the sample design, the dietary intake methodology used,
the wording of questions, the low completion rates and level of under-
reporting are among the limitations which were considered in conducting
this secondary analysis of the NFCS. These limitations were considered in
the selection of the study population, construction of variables and the

subsequent interpretation of results from analyses.

24

Seasonality

 

The original sampling scheme for the NFCS called for four random
subsamples to be executed over an entire calender year to permit control
for seasonal variation of food and nutrient intake. Crocetti and Guthrie
(1982) investigated seasonal variation utilizing the NFCS data set. They
explored all four quarters in terms of variation in meal and snack
patterns; variation in intake for major groupings of foods; and variation
in intake of individual nutrients and found no seasonal variation. In
view: of the evidence supplied by Crocetti and Guthrie (1982) that
seasonality did not have a substantial effect on the eating behavior and
nutrient intake of the study population, it was decided to merge the four
quarters and analyze the study population as an aggregate of the entire

year.

Phase I: Methods

 

The first task in Phase I was the basic selection and exclusion of
individuals to be included in analyses. Once the sample was selected
respondents were categorized by the primary researcher into the following
groups:

NON-MEDICAL respondents who identified one or more of the factors
affecting what they eat or drink and/or identified
themselves as being on a group or individual special diet.
The only factor excluded from the multiple item list of
factors affecting what a person eats or drinks in this
category was: "I have a medical problem ... ."

 

NON-MEDICAL respondents who identified one or more of the non-medical

E MEDICAE factors described above and identified one or more medical
problans. A medical problem included being on a doctor
prescribed diet and/or having a problem like diabetes or
an allergy.

MEDICAL respondents who identified one or both of the medical
problems.
NONE respondents who did not fit in any of the above categories.

These were the individuals who responded negatively or had
no answers to the two questions described above.

25

The essence of the first objective of this study was to determine if
persons who self-reported some non-medical or medical problems had
identifiable differences in terms of lifestyle or food and nutrient intake
from those persons who did not identify problems. These groups were
derived to aid in characterizing the Inajor descriptive differences
between NFCS respondents who reported no problems or factors affecting
their intake from those respondents who did identify factors or problems.

The next sections describe the variables used in Phase I to
characterize the four groups in terms of their: (1) socio-economic
characteristics; (2) meal and snack patterns; (3) nutrient intake; and (4)
food group intake and personal behaviors. In some cases variables were
used as they had been directly coded from the NFCS questionnaires. In
other instances variables were defined in terms of frequency or usage
reported in the study or reconstructed to incorporate information from
several response categories. An attempt was made to develop variables
which would be useful to nutrition educators and relevant to social,

demographic or lifestyle patterns.

Socio-Economic Characteristics

Socio-economic variables were used to differentiate the demographic
characteristics of the groups. Age was calculated based on the date of
birth indicated in the questionnaires. This information was used to
derive census age category groupings and RDA age/sex groupings. Sex, race
and region were used as recorded on the interview schedules. Many other
variables such as household size, family composition, respondent's
relationship to the head of household, poverty index, urban status,
working status and type of head of household, and education of head of

household were constructed either by condensing categories from the

26

questionnaire or incorporating two or more variables in a composite

variable.

Meal and Snack Patterns

NFCS respondents were asked to indicate the name of every eating
occasion recorded. The NFCS coding categories included: breakfast;
brunch; lunch; dinner; supper; beverage (coffee break); and snack. In the
initial data examination a frequency count of all patterns of meals and
snacks was computed for the three days. Two problems were encountered:
(1) there were a large number of eating occasions left uncoded on the USDA
micro data tapes and (2) there were a large number of individuals
reporting four or more meals per day. For various reasons (e.g.
respondents employed a term or name for the eating occasion not included
in the list) many of the eating occasions were left uncoded. Crocetti and
Guthrie (1982) recoded when possible the food intake occasions which had
been left uncoded by USDA employing very stringent criteria (see Crocetti
and Guthrie (1982) for criterion used). These recoded categories were
used in Phase I. This resulted in 13 percent of the individuals having
changes made in their labels for food intake occasions.

In further examination of the "four or more“ meals problem, it was
determined that many of the meals consisted of "multiple" breakfasts or
dinners recorded as taking place at different times, but within 1 to 1%
hours of each other. It was decided to consider breakfasts within 1%
hours of each other and all other meals within 1 hour of each other as
single meals.

After correcting for the two aforementioned problems, frequency
distributions were run listing the patterns of all possible combinations

of meals for the three days and the total number of snack events. The most

27

frequent patterns of meals were 333 (57%), 332 (17%), 322 (8%), 222 (7%).
The labels 332 and 322 did not indicate the sequence of days with three
and two meals. For example, 332 merely indicated that on two days three
meals were recorded and on one day two meals were reported. All records
not falling into one of the four major meal pattern categories were placed
in an "other" category.

Snack events consisted of either snack and/or beverage labels
reported with the same time label. In analyses, two Inajor snack
identification groups were used: any mention of a snack intake (snacker)
and no mention of a snack intake (non-snacker).

For all analyses, meal pattern categories included: 333, 332, 322,
222, and "other". Snack patterns were reported in terms of snackers and

non-snackers.

Nutrient Quality Assessment

 

Nutrient intake quality was the third area in which the study
population was examined. Nutrient quality data was obtained from the
three-day food intake records of respondents (it did not include vitamin
and mineral supplement data). A composite index was constructed from the
combination of the marginality score and the protein, fats, and
carbohydrates (PFC) score to assess nutrient quality. The marginality
score (MS) was constructed to give information about the number of
micronutrients that were ingested at a level designated as "marginal“.
Marginal for this investigation was determined at less than or equal to
59.9 percent of the RDA for that particular nutrient and age/sex category
of the respondent. The 59.9 percent for the RDA cutpoint was determined
with approval of the USDA project officer and a group of nutrition

professional consultants. This cutpoint was determined based on the

28

frequency distribution of respondents in age/sex categories and percent
RDA intake for nutrient and to reflect a reasonable cutpoint for
"marginal" for all micronutrients. The "scored" was a constructed value.
Worksheet #1 (Appendix A-I) illustrates the steps used in developing the
scoring system for seven nutrients. Crocetti and Guthrie (1981 c. and
1982) elaborate on the development of this index. Each of the seven
nutrients were assigned a weight according to the percent RDA achieved,
and the seven weights were summed to yield scores. The individual scores
ranged from seven (greater than or equal to 80 percent of the RDA for each
of the seven nutrients) to 63 (those cases where each of the seven
nutrients averaged less than or equal to 59.9 percent of the RDA).

The PFC score was developed to address the total caloric intake of
respondents in the proportions of protein, fats, and carbohydrates (PFC).
The PFC score method of development is shown in Worksheet #2 (Appendix A-
2) and further discussed by Crocetti and Guthrie (1981 b. and c. and
1982). The recent emphasis on the distribution of these three
macronutrients in terms of caloric intake calls attention to this
dimension of nutrient quality assessment not specified in the RDA. It
should be pointed out that there is neither firm basis nor authority for
setting the cut points as shown.

The PFC score was a constructed value in which individuals are
assigned scores determined by the patterns found for the proportions of
intake of protein, fats, and carbohydrates. A score which was identified
in the "okay" category indicated that each of the three micronutrients
were within the specified ranges indicated in Appendix A-2. All other
categories were determined by the number and specific macronutrient which

fell outside of the specified ”okay" percent intake categories.

29

A single composite index was constructed to explore the nutritional
quality of the diet of the respondents which combined the MS and PFC
scores. Several combinations were examined (Crocetti and Guthrie, 1982),

and the final 12 category composite is identified below.

12 Composite Index Categories*

 

 

 

 

PFC Index

Number of ONE TWO or MORE
Marginal All "okay" Outside Range OutSlde R3099
Nutrients

NONE B C

ONE 0 E F

TWO G H I

THREE - SEVEN J K L

 

*See Appendix A-5 for detailed descriptions of each
category A - L.

Each cell 'hi the above table represented one category group used in.
nutrient intake analyses. For example, the ll cell (NONE/ALL “OKAY")
represented those individuals who had no micronutrients equal to or below
59.9 percent RDA for the three day average for the seven MS nutrients and
their PFC ratio was all in the specified "okay" ranges. Those respondents
in the L cell (THREE-SEVEN/TWO OR MORE OUTSIDE OF RANGE) were those
individuals that had three or more nutrients that were equal to or below
59.9 percent for the three-day average RDA for the seven nutrients. This
group also had two or more macronutrients falling outside of the specified
ranges in the PFC score. This composite index was used to assess nutrient

quality in analyses.

30

Food Group Intake and Personal Behaviors

The major task in developing variables to assess food group intake
patterns lies in establishing and defining category groups. Crocetti and
Guthrie (1981 b.) defined two types of groups, one based on nutrient
content of foods and the other based more on usage and function of food
products in the U.S. diet. Crocetti (Crocetti and Guthrie, 1982)
developed food group categories primarily to reflect usage and function of
food items in the U.S. diet rather than nutrient content of foods.
Thirty-two groups were determined based on: function and content of
ingredients as they are marketed and/or as they appear or are used in meal
planning 'hi commercial restaurants or 'hi households' food selection;
major food subgroups coded by the USDA; and the frequency that respondents
mentioned specific food items (Appendix A-3 is a listing of the 32
categories).

Guthrie (Crocetti and Guthrie, 1981 a.) developed 13 groups to assess
food consumption in terms of nutrient contribution to the diet of
respondents. It was a composite of the Crocetti (Crocetti and Guthrie,
1981 a.) 32 food groups. Given the objectives of Phase I only the
Crocetti groups were used for analyses (see Appendix 3).

Several other variables were developed which combined data obtained
in the NFCS questionnaires. These variables were constructed to
characterize: eating outside the household; eating alone; use of vitamin
and mineral supplements; and weight status. Weight status was constructed
from the self-reported height and weight data obtained from respondents in
the questionnaires. From this information provided, the following

variable was used to compare weight status among respondents:

31

”Underweight", "Normal”, and ”Overweight" Status

"Underweight" = :5-15% of RDA range mid-points

”Normal" = -14.9% to +24.9% of RDA range mid-
points

”Overweight" = 2 +25% of RDA range midpoints
The values used to derive the cutpoints were from the ranges
recommended by the Food and Nutrition Board, Recommended Dietary
Allowances Committee (1980). The mid-points in the ranges reflect sex/age
groups for individuals less than 18 years of age and height groups for
persons 18 years of age and older. The percentage cut-points used for the
under and overweight categories vary due to the documented greater

potential physiological hazards of being underweight.

Analysis Phase I

 

Once the variables for Phase I were developed as described in the
previous four sections, frequency distributions for the variables were
cross tabulated with the medical and non-medical factor intake category
groups. Cambridge Computer Associates (1967) statistical computer
package, "Crosstabs," was used to characterize individuals with self-
reported problems. Contingency coefficients were estimated as the

measure of association between category groups used.

Phase II: Methods

 

The sample selected for analysis in Phase I was also used as the
study population for Phase II. The first task of Phase II was to describe
and select indicator variables to be used in the model presented in
Chapter II. The three sets of independent variables included were: (I)
demographic characteristics; (2) personal and food related behaviors; and

(3) nutrient intakes. The analytic procedures to derive indicator

32

variables for each of the sets and to estimate the relationship between
the three sets and the dependent variable (problem versus no problem
group) are described in this section. Discriminant analysis, factor
analysis and canonical correlation were all applied in deriving the
variables to be used in the independent variable sets and in estimating
relationships between variables in the model.

A data tape was obtained from the Crocetti and Guthrie (1982)
secondary analysis of the NFCS. It contained data on variables as they
had been coded by USDA on the household and individual questionnaires. It
also contained variables that were constructed for analysis in Phase I

(See Chapter III, Section II).

Description and Transformation of Variables

The first task in Phase II was to describe all variables included on
the tape obtained from Crocetti and Guthrie (1982). A 10 percent random
subsample was generated from the study population. SPSS, Statistical
Package for Social Sciences (Nie et al., 1975) Subprogram on Frequecies
was used to examine missing values and frequency' distributions of
dichotomous and polychotomous variables. Multivariate analysis
techniques used in this investigation assume linearity1 among the
variables. As Green and Carroll (1976:7) indicated:

The assumption of linearity in the parameters, is not
nearly so restrictive as it may seem. First, various
preliminary transformations (e.g., square root,
logarithmic) of the data are possible in order to
achieve linearity in the parameters. Second, the use
of "dummy" variables, coded, for example, as
elementary polynomial functions of the "real"
variables, or indicating category membership by
patterns of zeroes and ones, will enable us to handle
certain types of nonlinear relationships within the
framework of a linear model.

33

The statistical techniques used in this study can utilize binary-coded
(zero-one) variables, but not multiple categorical variables. Therefore,
all polychotomous variables were recoded as a set of dichotomous "dummy"
variables.2

The SPSS (Nie et al., 1975) Subprograms Condescriptives and Oneway
produced information on the number of missing values, and on the means,
standard deviations, ranges, skewness, and tests for’ homogeneity' of
variance for all non-categorical independent variables. Beyond assuming
relationships between parameters were linear, using general linear model
(GLM) multivariate analysis techniques for statistical testing and
parameter estimation it is "necessary to make various assumptions in the
form of restrictions on the model and the error term Ut“ (Hanuskek and
Jackson, 1977:47). Hanuskek and Jackson (1977:47-50) continue by

summarizing these assumptions:

1. l r23|.:1.0 The explanatory variables may be correlated,
but not linear dependent.

2. Fixed X. The correlation/covariance between exogenous
(independent) variables and error of prediction = zero.

3. E(Ut-U) = (I All observations on the dependent variable
implicity include an unmeasurable error term.

4. (a) nut-U)2 = 02,

02 for all t.
(b) Hut-U) (US-II) ots = o for t f s

All possible error terms associated with one observation
are independent of, and thus uncorrelated with, the error
terms at other observations (homoskedasticity) and ‘the
error of different observations do not covary (non-
autocorrelated).

Monge (1980:24) states that "if these assumptions are warrented,
then statistical theory regarding sampling distributions and properties
of estimators can be used to formulate inferences about the parameters."

The application of the nmltivariate techniques for this investigation

34

varied in their purposes and therefore the violation of the assumptions
would vary in the affect on the results. For instance, factor analyses
were applied for exploratory purposes, to explore and detect patterns in
the variables. Factor analyses were not used for purposes of statistical
testing or parameter estimation. Therefore, the violation of these
assumptions was secondary. In general the multivariate analysis
techniques were not specifically for purposes of statistical testing. The
techniques were applied for parameter estimation primarily. This is a
critical point in the examination of GLM assumptions and assumptions
specific for various analyses applied. The violation of the assumptions
becomes secondary. In each section the criteria for assumption
examination and violation is specified.

The assumptions related to normal distribution, homoskedasticity,
and non-autocorrelation were explored for the independent variables when
appropriate. It was assumed that if the independent variables were
homoskedastic and non-autocorrelated, then the error terms would also be
found in these forms. Large positive skewness was used to indicate that
the independent variables were gpt_normally distributed about the mean.
Measures of homogeneity of variance, homoskedasticity, were obtained from
analyses of variance (ANOVA) with the problem versus no problem group
variable. Variables which were highly skewed 35I3.0| (Stoyanoff, 1981)
and/or had a maximum/minimum ratio 3.|2.Ol were transformed to correct for
non-normal distributions and heteroskedasticity. Single bend
transformations, such as square root and logarithmic, were explored for
the variables not meeting the two aforementioned assumptions. The SPSS
Subprograms Condescriptives and Oneways were rerun on the final
linearized functional forms obtained for transformed variables. The

transformed variables were used in all further analyses.

35

Tests for Multicollinearity

The second task in Phase II was to determine those variables for
inclusion in the model representing the three sets of independent
variables. Discriminant Analysis and Factor Analysis SPSS Subprograms
(Nie et al., 1975) included the statistical techniques used to make these
decisions. From the 10 percent random subsample generated from the
initial analyses in Phase II, two random subsamples were determined. Each
of these subsamples represented 50 percent of the 10 percent initial
subsample (or each of the subsamples reflected five percent subsamples of
the total study population N = 24,362). The two subsamples were generated
for purposes of cross validation.

Criteria was determined to aid in making decisions regarding
inclusion of descriptive variables for each of the three independent sets
in the model as specified in Figure 1, Chapter II. The initial decision
making criterion for exclusion of variables included redundancy' of
information used in developing or constructing variables. For example,
the Crocetti 32 food groups and the Guthrie 13 food groups (Crocetti and
Guthrie, 1981 b.) were created from the same respondent data. Therefore,
only one set was used in the analyses for Phase II. The second criterion
applied was a series of variables adding up to a composite total. An
example of this would be the three variables representing the three day
average percent intake of protein, fat and carbohydrate which added up to
a composite 100 percent. All three variables could not be used in the
analyses because they were linearly dependent which violates the basic
assumption of the GLM . Consequently, only two of the three were used in
the model. These two criteria eliminated several potential indicator
variables from inclusion in the variable sets within the model.

The goal of the next series of computer runs was to derive the three

36

sets of variables which reflected linear independence. Multicollinearity
is the condition in which one or more of the variables included in a
sample of data are linearly dependent (Hanushek and Jackson, 1977).
Severe multicollinearity has the effect of making the determinant of the
correlation matrix approach zero. When linear dependency exists the
dimensionality of the space needed to represent the variable vector is
less than the number of variables.

Multicollinearity reflects some level of covariation among
independent variables which may have the following implications in
relation to multiple regression (Neter and Wasserman, 1974:344):

(1) The estimated coefficients tend to be quite
imprecise and the true coefficients tend to lose
their meaning. '

(2) The coefficients of partial correlation between
the dependent variable and leach of the
independent variables tend to become erratic
from sample to sample.

(3) As the correlations between the exogenous
variables increases, the variance between the

estimated coefficients becomes larger.

(4) Standard errors of coefficients will tend to get
larger.

SPSS Factor Analysis Subprogram (Nie et al., 1975) was used to: (I)
obtain correlation matrices for the three sets; (2) calculate the
determinant of the correlation matrices; and (3) derive rotated factor
matrices to demonstrate independence among the variables included in each
set. The rotated factor matrices showed that each of the major factors or
axes determined in the factor eigenanalyses were primarily explained by
gag of the indicators in the set. If several indicators had mutually
explained the major factors or axes linear dependency would have been

exhibited in the sets of variables.

To check for linear independence among indicator variables and make

37

decisions regarding the exclusion of variables which contribute to the
dependency, correlation matrices and their determinants were generated
for the three sets of independent variables. The end result of this
series of computer runs was a set of relatively independent variables
representing each of the three composite theoretical factor groups which
affect food related behaviors. "Relatively independent“ was quantified
by determinants of correlation matrices greater than zero, positive log
determinants of covariance matrices, and major factors or axes determined
in factor eigenanalyses primarily explained by ggg of the indicators in

the variable sets.

Discriminant Analysis

 

Three discriminant analyses via SPSS Subprogram Discriminant (Nie et
al., 1975) produced the correlational relationships between the dependent
variable, the problem versus no problem groups, and the three sets of
independent variables: (1) demographic characteristics; (2) personal and
food related behaviors; and (3) nutrient intake. Discriminant analysis
may be treated as a special case of canonical correlation analysis, the
general procedure for investigating the relationship between two sets of
variables (Knapp, 1978). As Knapp (1978:414, 415) stated:

Discriminant analysis is actually a multivariate
analysis of variance in reverse, that is, there is
one categorical dependent variable of group
membership and there are two or more continuous
independent variables. The p independent variables
are treated as though one were carrying out a
multiple regression analysis, and the dependent
variable of k categories is coded in the same way that
an independent variable is treated in one-way
analysis of variance, that is, by creating q = k-1
dummy dichotomies. A standard p x q canonical
analysis is then applied to the resulting system, and
the F test of the largest canonical correlation
coefficient determines whether or not the k groups
are significantly separable on the p variables.

38

The specific underlying assumptions for statistical tests (as oposed
to estimation alone) of discriminant analyses are summarized by
McLaughlin (1980:178):

(1) Multivariate normality-p independent variables
have multivariate normal distributions in each
of the populations from which the k groups are
sampled.

(2) Homogeneity of the population dispersions-
population variance-covariance matrices are
equal (common covariance matrices).

(3) Mutual exclusivity and exhaustiveness of k
groups.

The seriousness of the violation of these assumptions was viewed in light
of the use for estimation rather than statistical testing. The
dichotomous (dummy coded) independent variables did not meet the first
assumption. The seriousness of this violation was determined through
evaluation of the second assumption. It was decided by the researchers
that if the covariance matrices log determinants were comparable for the
problem versus no problem groups for each of the sets of indicator
variables than discriminant analyses could be applied without serious
consequences to the estimated parameters. Therefore, the criterion of
comparable covariance matrices was used to evaluate the consequences of

violation of the underlying assumptions of discriminant analyses.

Canonical Analysis

The final task in Phase II determined the correlational
relationships between the three sets of independent variables. SPSS,
Canonical Correlation (Nie et al., 1975) and BMDP, Canonical Correlation
Analysis (Dixon and Brom, 1979) produced canonical correlations and
canonical variable loadings for each canonical variate accounting for

large proportions of the variance in the observed variables. In canonical

39

analysis, the reference axes determined are called canonical variates.
The canonical variates represented mathematical structures derived from
eigenanalyses. Canonical Variate 1 for example is the first reference
axis derived and accounts for the greatest proportion of variance between
the two sets of observed variables. The canonical variable loadings
helped identify those specific variables which accounted for the
proportion of variance explained by each of the canonical variates. The

results of these analyses are discussed in Chapter V.

4o

FOOTNOTES

1A model is called linear (or linear in the parameters) when the

effects of the various predictor variables are treated as additive.
In the expression y = blx1 +bzx2 +....bnxn, y is composed of a linear
combination of variables and regression parameters, the parameters
being each of the first degree (Monge, 1980; Green and Carroll,

1976).

Green and Carroll (1976:8-9) and Nie et al. (1975:374-375) describe

how the recoding is done.

CHAPTER IV
RESULTS AND DISCUSSION: PHASE I

The findings from Phase I of this investigation are presented in this
chapter. The selection of the study population from the 1977-78
Nationwide Food Consumption Survey (NFCS) and the factor intake category
groups by which the study population was analyzed are described. The
factor intake category groups were determined based on the self-reported
factors and problems respondents identified as taffecting their 'food
consumption. Results from cross tabulations between the factor intake
categories and each of the following: meal and snack patterns, socio-
economic characteristics, nutrient quality assessment, and food intake
and personal behaviors are discussed. The results from cross tabulations
build upon cross tabulations in previous sections. The end product is a

composite typological description of each factor intake category group.

The Sample

 

The first task in this analysis was the selection from the NFCS data
tapes of individuals for the study population. The following criteria
were used sequentially in the deletion process: (1) less than four years
of age; (2) no age; (3) no race; (4) reported pregnant/nursing; (5) diet
record for one day only; (6) diet record for two days only; (7) no
relation to the household; and (8) "other race". Table 1 displays the

sequential deletion of individuals to determine the 24,362 respondents

41

‘42

.m_m=uw>_u:_ yo Logan: m=_ucmum we ucwucoa p.o can» mmmg +
.m_m=u_>wu:w we Lassa: acwucoum co woman ucwucom «

 

 

 

A_.oav Am.m~v Ae.m~v Ac.mav A~.omv .R
News 98 mote :3. SN m8 ~86 98 "8; 21m ...: :3 N8; «.2 8%; SEE
Nom.8~ m.~ new eao.e N.~ oHN mmo.c m.m NAN ma_.m m.~ mo" Nae.“ m.~ mHN moan =ngbo=
www.mm _._ mam emm.m o._ on mom.o m._ com mmm.m P._ NA .om.~ _._ so” u_o;mm=oz 8:3
cu tmuo—mgcz
cam.m~ ¢.~ NMN oom.o N.N HRH mo¢.o m._ mefi o_¢.m ..N mm“ _Q¢.~ ~.m Hmm xpco ago N
"Ugoumm um;
mom.om a.m on.H .mm.m o.o owe omm.o ~.m mme mqm.m «.8 mme Nae.“ R.m com »_=o sac _
”ucoomm umwa
mmp.m~ o._ mow .mm.o m.o No mmm.o m.o mm Noa.m 0.. mm NmF.m N._ OHS m=_oabuaa\u=a=maca
~N¢.m~ ~.o So awe.“ P.o SH Ago.“ _.o N omo.o ~.o m" Nm~.m m.o on cobabm Guam oz
mm¢.m~ ..o cm moo.“ + e cmo.m F.o HS meo.c + N NNm.m + m umoaom ma< oz
wom.m~ ¢.~ ~o~.~ mac.“ m.m amo moo.N ~.m "no mvo.o N.m mmm mmm.m ~.m Mme mums Lo
u_o mama» a
cucumm .R * =_aoam .N a =_aomm .N a :_aoam .& . =_abam .N ‘ =o_oa_uo
Low mcommmm
oLN.oMuz ~H~.~uz coo.anz vwm.o "z m-.muz
Hmmmw “mmmflm Hamm umaamw mmflumw

 

 

muuz mm--o_ mo Loo» .muo» use mcmucmso Laou Lo» umm mama EoLw mucmvcoqmmm eo cowum—mo

_awo=a=cam ._ m_nah

43

used for this investigation.

The first group deleted were infants and children less than four
years of age (7% of the total population); they presented analytic
problems since the nutrient needs and eating patterns of these age groups
vary markedly from adult patterns. Additionally, these respondents were
unlikely to have filled out their diet records. Therefore, the standard
methodologic formats or indices used would not have adequately
accommodated these respondents.

The next two groups deleted were those persons with no age or race
stated. These groups were small in number; and age and race were used
extensively throughout this analysis.

Pregnant and lactating women were then deleted due to the specialized
nutrient needs and eating patterns of this group. This group had only a
few individuals (about 1% of the total population).

The largest number of deletions (8% of the sequentially deleted
population) included all those cases without a complete three day food
intake record. Retention posed analytic problems without satisfactory
solutions. Treating them as a separate analytic group was considered.
The sequence of days within the three requested showed numerous and varied
permutations which limited the possibility for grouping even further.
Additionally, the significance of these variations was not clear. .A
weighting scheme was considered however, there was no theoretical ground
for any specific weighting.

The final two categories omitted in the sequential deletion process
from the four quarters and the total year were unrelated individual
members in the household and the "other race" category. ‘Three hundred and
forty-eight persons (1%) represented roomers or boarders in the

households interviewed in the survey. Socio-economic variables used in

44

analyses of the study population were described by the head(s) of the
household attributes. To characterize roomers and boarders by an
unrelated head(s) of households' socio-economic status would have been
inaccurate.

The undefined category of "other race" included 866 persons (3%) of
the study population, once the first seven categories for deletion had
been applied. Race was used as a discriminating variable in many analyses
throughout the investigation. “Other race" did not include enough
individuals for meaningful analysis compared with the race categories:
"white" and "black."

The cases chosen for deletion were examined to see if they were
random and if deletion would further bias analyses of the population
retained. From Crocetti and Guthrie's (1982) tables representing the
distribution of deletion categories by sex/race/age, it was determined
that the deletions resulted in random and approximately unbiased

rejection of similar proportions of individuals from each quarter.

Factor Intake Categories

Factor intake category groups were derived as illustrated in Figure 2
from the 24,362 respondents (see Table 1). Approximately 50 percent of
the persons included in this investigation reported some factor which
potentially affected their food consumption. The largest group (39.5%)
included respondents who identified one or more of the factors in the
question: ”these are things which may affect the way you eat or drink"
(excluding the medical problem) or reported being on a group or individual
special diet. Ten percent reported a medical factor, of which, 60 percent
(1,454 respondents) also identified at least one non-medical factor. The

12,308 (50.5%) respondents who did not fit into one or more of these

45

cowunE=mcou woo; Lows» uumw$< an: cows: mcouoou

_oo_ooz-ooz ooo zoooooz ooozooooooo moooooom ooooo_oooo mouz o~--o_

 

Azm.om .mom.~_ u zv Aoo.o .ooo u zv zzm.om .oNo.o u zv

mzoz "mcouuoo mzouco o: mcouuou mcouoou 4<uHomz

oo_ooo=ooo ozz mooocoa ooozooz zoo; moomcoo -zoz zoo; moomcoo
oz mmz oz

 

.N mmonm

Azo.o .omo.z u zv

mgouumu Aqumomz a
4<uHomzuzoz cuwz mcomgma

mm>

 

Amcouuou PoUwumz noncommcum_mmv
”Loozou

“om.om .oom.m_ u zv Rom.oo .oao.__
.mm WMH

Fl ................

AmL0puom Pouonmzucoz cmocommguw—mmv
“Loo_ou

Amom.omnzv momm—oco so vows—ucw mcomcmo

"Lomz Pouch

u zv

46

categories acted as the control group in the study. The four category
groups were combined into one variable to cross tabulate with the
variables used to identify and characterize meal and snack patterns,
socio-economic descriptors, nutrient quality and food intake and personal
behaviors. Each group description builds upon descriptions in previous
sections. The end product is a typology of each group.

1 (C.C.) was used as a measure of

The contingency coefficient
association. It was based upon chi-square and takes the N of the sample
into consideration. Due to the large sample size (large N's), the
statistical tests applied to assess if systematic relationships existed
between the categories always indicated low or no association between
groups (contingency coefficients 5 .3 and chi-square p's _<_ .001).
Therefore, the contingency coefficients and the significance level of all

reported tables in this chapter were C.C. <_ .3 and p: .001 unless

otherwise noted.

NON-MEDICAL Category

 

To identify the specific factors reported as affecting the way the
respondents in the NON-MEDICAL category ate or drank, Table 2 was
produced. Table 2 represents the percent distribution of respondents with
NON-MEDICAL intake factors by number and type of factor affecting intake.
"Don't like certain foods" (35%), and "I don't feel like eating breakfast
early in the morning" (23%) were the items most often mentioned by the
total group (N = 9,620). "I'm on a diet to put weight on" (1%), "I have a
chewing problem because of teeth" (3%), being on a group or individual
special diet (6%), and "I have no interest in cooking for one" (6%) were
the items respondents least often mentioned. Sixty-seven percent of the

respondents mentioned one non-medical factor. Two, three and four to six

 

47

 

 

o~o.m co. ~.¢ ~.em ~.o o.m~ c.o_ —.m m.m m.o ~.p_ u 4<ho»
mm_ oo. ~.m m.mp m.~— o.m— N.mp e.m c.m o._ ~.ep xmmumzou
.mo oop m.~ ~.~N m.o_ ~.m~ o.¢— o.e ~.w e.o m.mp ummxh
mm~.~ co. m.m N.m~ m.o o.m~ m.NF ¢.~ m.o_ m.o m.mp ozp
omo.m oop p.m m.oo m.m ~.m~ ~.~ o.~ m.— m.o e.m mzo
“mucouuou
* a Loguo muoou _ Low umooxomcm mmcm< Em_aoco um_c _o_umam cwoo owe; —ouonmxicoz
cpoucwu mczxoou oJPA u.=oo o.:oc o=_:m:u pooowmzncoz cu umwo cu um_o mo consaz
mx_4 u.=oo oxwa u.coo mnoou
4<pop a Leuuou ~ouwow:-:oz to max»

 

 

Aomo.o.zv ozoooz oozoooooz Avaoooou
_oooooz-zoz to Loooou to oozo ooo LooEoz zo mtoooou ozoooz ozozooz-zoz zoo; moooooooooz Lo oozooooooooo ooootoo .N o_ooz

48

factors were mentioned by 24 percent, seven percent, and two percent,
respectively, of the population segment.

Table 3 has the distribution of respondents with NON-MEDICAL intake
factors by sex and race and number of reported factors. All sex/race
categories had more than 62 percent of the respondents with one problem.
The black males had the most (80%) and the white females the least (62%)
with one problem reported. The white females most often report two or
more factors. In general, the males and blacks identified fewer multiple

item responses than the females and whites.

NON-MEDICAL AND MEDICAL Category

 

Table 4 has the percent distribution of respondents with NON-MEDICAL
& MEDICAL intake factors by the number and type of non-medical factors
reported. In this group (N = 1,454), "Some foods don't agree with me"
(24%), "I'm on a diet to lose weight" (20%), and ”I don't like certain
foods" (19%) were the most frequently reported non-medical factors. The
distribution of respondents in the number of non-medical factors reported
was: one (56%), two (27%), three (12%), and four to seven (6%). This
category group seemed to have reported more multiple item responses than
the NON-MEDICAL only group (Table 5). The two medical factors were
identified by approximately the same percentages within the one and two
problem category groups (Table 6).

Table 7 represents the percent distribution of respondents with
reported NON-MEDICAL & MEDICAL intake factors by sex and race and type of
non-medical factor. Males had more responses in the "I do not like
certain foods" than the women; women had more in the "I am on a diet to
lose weight" category than men. Women were found (distributed by RDA

age/sex categories) to begin reporting being on diets to lose weight in

49

Table 3. Percent Distribution of Respondents with NON-MEDICAL Intake
Factors by Sex and Race and Number of Factors Affecting
Intake (N = 9,620)

 

 

 

 

 

 

Number of Factors % TOTAL

Sex and 1 2 3 4-6 % #
Race
MALE

White 72.0 21.7 5.0 1.3 100 3,506

Black 80.0 16.3 3.2 0.4 100 465
FEMALE

White 62.4 26.2 8.7 2.8 100 4,957

Black 66.9 23.1 7.5 2.4 100 692
TOTAL 67.1 23.9 7.0 2.1 100 9,620
MALE 73.0 21.1 4.8 1.2 100 3.971
FEMALE 63.0 25.8 8.5 2.7 100 5,649
WHITE 66.4 24.3 7.1 2.1 100 8.463
BLACK 72.2 20.4 5.8 1.6 100 1,157

TOTAL 67.1 23.9 7.0 2.1 100 9,620

 

5C)

 

 

 

omo._ oo_ ~.o o.o_ o.o o.o_ m.m~ o.m ~.o _._ m.o~ a o<ooz
.m oop o.o ~.o~ o.~_ ..om m.o_ o.o o.~ m.~ ~.m_ zuzmm-z=ou
Roz ooz o.~ m.o~ o.o o.z_ o.oN o.~ o.m o.o ~.op mozzh
mom oo_ o.m o._~ o.o m.o_ o.mN m.m m.o m._ _.N_ oz»
o_m oo_ m.o ~.o_ ~.N m.o o.m~ o.m ~.m o.o o.om uzo
ASSN—uni
z u cacao mvoou P Lo» umowxoocm omcm< Ewpnoca uw_o _o_uwam zoom moo; Foupeoz-coz
oooocoo ooozooo ozoo o.ooo o.ooo oo_:ozo .oooooz-ooz oo oo_o oo oooo to Loosoz
oz_4 o.ooo oz_o o.ooo mooou
Dzooz o Loooou .oooooz-ooz Lo ooze

 

 

Aooo._ . zv ozooo_ ooooooooz Aooaoooou
.oooooz-ooz co oozo ooo Loosoz zo otoooou ozoooo ozooooz o o<o_ooz-zoz zoo: mooooooomoz Lo oooooooaoooo ooooooo .o o_ooo

51

 

 

 

 

 

 

 

 

Table 5. Percent Distribution of Respondents with NON-MEDICAL 8
MEDICAL Intake Factors by Number of Non-Medical Factor(s)
Affecting Intake (N = 1,454)
Number of
Non-Medical
Factor(s) %
ONE 56.3
TWO 26.7
THREE 11.5
FOUR-SEVEN 5.6
TOTAL 100.0
Table 6. Percent Distribution of Respondents with NON-MEDICAL 8
MEDICAL Intake Factor(s) by Number and Type of Medical
Problem (N = 1,454)
Type of Medical Problem % TOTAL
Number of Medical Problem Dr. Prescribed
Medical Problem(s) Like Diabetes Special Diet % #
ONE 59.4 59.3 51.0 1,083
TWO 40.6 40.7 49.0 1,042
TOTAL % 100.0 100.0 100.0

 

52

 

 

 

 

 

oo_ oo_ o.o o.o_ o.~ ~.o_ ~.o~ o.o o.~ o.z ..om zo<zo
oo~._ oop o.o N.op ..z o.o. o.oN o.o o.o P._ o.oN “Caz:
moo oo_ o.o o.o. o.“ N.m_ ~.m~ o.o N.m P._ _.- mo<zuo
ooo oo_ o.o “.mm o.o ..N. o.o~ o.o o.o o._ o.o. “Dz:
omo._ oo. N.o o.o_ o.o o.o. o.m~ o.o z.o _._ o.om ozoo»
zz_ oo. o.o ~.oz m._ m.~, ..mN ..o o.o _._ ~.o~ zoo_o-o_osou
.oo oo. o.o ~.o_ o.o o.o. N.mN z.m o.o _._ m._~ oo.z3-o_oeou
mo ooz ~.~ N.oN o.o o.N_ m._o o.o o.o o.o o.o. zoozo-o_oz
moo oo_ o.o o.m~ o.o o.~_ ~.om ..o A.” z._ ~.op ooozz-o_oz
z a tango muoou p so» umooxomcm mocm< Em_nozo um_o _m_owom c_aa mmoo woom\xmm
oooocoo ooozooo ozoo o.ooo o.ooo oo_:ozo _oo_ooz-ooz oo oo_o oo oo_o
oz_o o.ooo ozoo o.ooo mooou
4<»o» a .ou_co:-coz mo max»

 

 

zooo._ . zo ozoooo oooooooc< Aooooooou
Poo_ooz-ooz Lo oozo ooo oooz\xom zo Amvtoooou oxoooz ozozooz o o<o_ooz-zoz zoo; moooooooooz Lo ooooooocoooo oooocoo .N o_ooo

53

the 11-15 age category.

MEDICAL Category

 

Table 8 shows the distribution of the respondents who identified a
MEDICAL intake factor by the number and type of factor. Of those
respondents (53%) who list one medical problem, 69 percent (357
individuals) identified it as being "a medical problem like diabetes."
Forty-seven percent of the total group (N = 980) self-reported both types

of medical problems.

Discussion: Intake Factor Category Groups

The U.S. population on special diets has been studied elsewhere. The
results are reported here for comparison with our results. The percent of
the population 12-74 years of age on a special diet were determined by
National Center for Health Statistics (NCHS, 1978) using the Health and
Nutrition Examination Survey (HANES) data. Approximately 11 percent of
the total population were on a special diet. The category "special diet."
was further described by reason for and kind of diet. The reasons
included: overweight, diabetes, ulcer, heart trouble or high blood
pressure, and other. The kinds of diets included: low fat, low salt, low
carbohydrate, low calorie and other. Approximately 40 percent of the
respondent's identified "overweight" as the reason for the diet and 34
percent identified "other". Forty-three, 39 and 38 percent respectively,
reported “other", "low calorie", and "low fat" as the kinds of diets.
More females (13%) than males (8%) reported being on a special diet. More
white (11%) than the black (9%) HANES respondents reported being on a
diet. Similarly to HANES our study population had more females than

males and more whites than blacks report being on a diet.

54

Table 8. Percent Distribution of Respondents with MEDICAL Intake
Factor(s) by Number and Type of Medical Problem (N = 980)

 

 

Type of Medical Problem % TOTAL

 

Number and Type of Medical Problem Doctor Prescribed

 

Medical Problem Like Diabetes Special Diet % #
1 - Medical Problem
Like Diabetes 43.6 -- 36.4 357
1 - Doctor Prescribed
Special Diet -- 25.8 16.4 161
2 - Medical Problem
and Doctor Pre-
scribed Diet 56.4 74.2 47.1 462
TOTAL % 100.0 100.0 100.0 980
TOTAL # 819 623

 

55

About 15 percent of the HANES respondents had a chewing problem
(NCHS, 1977). The NCHS report also included 1974 Health Interview Survey
weight control data among persons 17 years of age and over. Of those
overweight persons approximately: 64 percent were trying to lose; 23
percent were trying to lose on physician's advice; and 57 percent
identified being on a diet (of any kind).

General comparisons of the NCHS and NFCS findings were attempted.
The total percent of persons identifying some type of special diet were
comparable in both studies, HANES (10%) and NFCS (approximately 10%). In
both HANES and NFCS more females and whites than males and blacks reported
being on special diets or having some problem with their intake,
respectively.

In general the NCHS (1977, 1978) findings and our results were
comparable regarding proportions of the population with factors which may
affect their intakes. The largest difference was the large percentage of
HANES respondents and small percentage of NFCS respondents identifying

chewing problems. This likely is due to data collection differences.

Identification of Meal and Snack Patterns, Socio-Economig,
Nutrient and Food Characteristics of Factor Groupings

In following sections, the four factor intake category groups were
crosstabulated with variables used to identify and characterize meal and
snack patterns, socio-economic descriptors, nutrient quality assessment,
and food intake and personal behaviors. Each is discussed separately;
however, understandings gained from the analyses with each of the
descriptor groupings built upon one another. Therefore, discussions of
each include findings from previous sections. The end result was a

composite typology of each of the four category groups: NONE, NON-

56

MEDICAL, NON-MEDICAL 8 MEDICAL, and MEDICAL. Whenever appropriate,
findings from this investigation are compared to Crocetti and Guthrie
(1982), other NFCS investigations, NCHS data, and other relevant

research.

Meal and Snack Patterns

It's generally believed that Americans no longer eat in a traditional
three meals a day pattern. The data obtained from the NFCS supports the
belief that large percentages of the U.S. population are not eating the
traditional three meals a day; and that snacks are being eaten.

The distribution of respondents by meal pattern and intake factor
category groups are illustrated in Table 9. The NONE and MEDICAL groups
had the most respondents in the 3,3,3 pattern (meal patterns are described
in Chapter III). Conversely, the two non-medical groups identifying some
type of factor had more individuals in the composite of the non-3,3,3
patterns (3,2,2, 2,2,2, and 'other' categories).

The investigation of snacking behaviors (Table 10) demonstrated that
each of the groups had significant percentages of respondents (73%)
reporting at least one snack in the three-day intake period. The groups
reporting non-medical problems had the largest percentages of respondents
reporting the consumption of one or more snacks. In crosstabulating
snacking behavior with meal patterns, the 2,2,2's had more respondents
that reported snacking in each of the factor intake categories. It was
the groups reporting non-medical problems that also had the largest
percentages of respondents in the non-3,3,3 meal patterns.

These findings agree with those of Crocetti and Guthrie (1982): the
fewer labeled meals consumed by respondents, the more snack events

reported. Nutrition educator's and marketer's strategies must

57

Table 9. Percent Distribution of Respondents by'Intake Factor Category
Groups and Meal Patterns ( N = 24,352)

 

 

 

 

Intake Factor Meal Patterns % TOTAL
Category Groups 3,3,3 3,3,2 3,2,2 2,2,2 Other #
NONE 63.4 16.3 6.9 5.4 8.0 12,308
NON-MEDICAL 48.9 17.0 10.5 10.0 13.6 9,620
NON-MEDICAL 8

MEDICAL 56.1 14.9 7.6 9.4 12.1 1,454
MEDICAL 66.7 16.4 4.5 5.7 6.6 980
TOTAL % 57.4 16.5 8.2 7.4 10.4
TOTAL # 13,986 4,010 2,009 1,813 2,544 24,362

 

Table 10. Percent Distribution of Respondents by Intake Factor Category
Groups and Snacking Behavior* (N = 24,362)

 

 

Snackinngehavior* %

 

Intake Factor

 

Category Groups Snacker Non-Snacker
NONE 74.3 25.7
NON-MEDICAL 80.7 19.3
NON-MEDICAL 8

MEDICAL 77.4 22.6
MEDICAL 73.0 27.0
TOTAL % 77.0 23.0

 

*
Snacking Behavior categorized respondents by whether or not they
identified consuming any 'snacks' in the 3 day intakes (Snacker)
or no intake occasions labeled 'snack' (Non-Snacker).

58

incorporate the realizations that the three meals a day pattern does exist
for some, Yet is supplemented with snacks. Additionally, large
percentages of the population are eating non-three meal a day patterns and

are also eating snacks.

Socio-Economic Descriptors

The distribution of respondents in the intake factor category groups
crosstabulated with socio-economic descriptors: census age categories;
size of household; family composition; relationship of respondent to head
of household; poverty level index; education of head of household; working
status of head(s) of household and four census regions are illustrated in
Tables 11-18.

The NONE category and the NON-MEDICAL category had more respondents:
less than 18 years of age; living in households sized four to 14; in
families composed of adults and children and teens; with above poverty
level incomes; and working male and female headed households. The NONE
category had more children ('< 18 years) than the NON-MEDICAL category.
And the NON-MEDICAL had more adults (25-55 years) and more respondents
with above poverty level incomes than the NONE group. Otherwise the
socio-economic descriptions of the two groups were similar. This may be
an important finding for nutrition education efforts generated for these
two population groups.

On the other hand, the categories which reported some kind of medical
problem, NON-MEDICAL 8 MEDICAL and MEDICAL, were similar' in socio-
economic composition. Each of these groups had more respondents in the
older age categories. The NON-MEDICAL 8 MEDICAL group had more in the 45-
75 years categories while the MEDICAL had the largest in the 55 years and

older census categories. These groups also had the largest percentages of

59

 

 

 

Nem.¢~ mom mem._ omm.~ o—¢.~ mmm.~ emo.m com mmm._ mem.~ owc.m omm ‘ 4<ho»
oo_ ~.m o.~ m.m m.a m.m w.m_ m.m m.@ m.—_ v.m~ v.~ a 4<hop
omm oo. “.mp _.m~ e.- o.p_ o.o o.o m.p m.— —.m e.m m.p 4<u_omz

omv._ cop o.m m.m_ o.om m.e_ ~.o_ ..m m.~ m.~ ~.e —.o o.o 4<u_omx a
4<u_omzuzoz
o~o.m oo— o.m ~.m m.m m.op o.—_ N.@. m.c _.~ ~.~— o.o. m._ 4<u~ouzuzoz
mom.~_ 00— ~.~ e.m N.“ ~.m o.o m.o_ m.m m.o o.mp o.om o.m uzoz
* a +m~ mmvume movumm mmvumc movumm mmv-mm mmvumm -v1m_ mpv-__ ppVTm e mesecw xcemeuou
Leuueu exeueu

4<h0h a mmcoeeece em< mamceu

 

 

Amom.v~n zv mmeoeeecc eo< memeeu ecu meeecw xceoeeou Leuueu exeu:_

xe mueeeceemem we :epuzewcumoo aceucea ._— epneh

60

Table 12. Percent Distribution of Respondents by Intake Factor Category
Groups and Size of Household (N = 24,353*)

 

 

 

 

 

Intake Factor Size of Household % TOTAL
Category Groups 1 2 3 4 5-14 % #
NONE 5.9 15.4 15.0 23.6 40.1 100 12,308
NON-MEDICAL 10.3 18.8 15.5 22.2 33.1 100 9,611*
NON-MEDICAL 8

MEDICAL 20.6 32.3 16.4 14.9 15.7 100 1,454
MEDICAL 18.8 36.8 14.3 14.5 15.6 100 980
TOTAL % 9.1 18.6 15.3 21.3 34.9 100
TOTAL # 2,197 4,531* 3,715 5.396 8.505 24,363*

 

*Excludes 9 individuals who had "no answer" to size of household.

Table 13. Percent Distribution of Respondents by Intake Factor Category
Groups and Family Composition (N=24,362)

 

 

 

 

Intake Family Composition % TOTAL
Factor
Category AdUltS AdUltS BRO AdUITS BHO Children
Groups Only Children<11 Teens 11-18 and Teens % #
NONE 26.9 22.5 22.2 28.5 100 12,308
NON-MEDICAL 33.9 21.6 21.8 22.7 100 9,620
NON-MEDICAL 8

MEDICAL 61.7 12.0 16.0 10.2 100 1.454
MEDICAL 65.1 11.3 12.9 10.7 100 980
TOTAL % 33.3 21.1 21.3 24.4 100

TOTAL # 8,102 5,135 5,188 5,937 24,362

 

61

 

Nom.¢m nvo.m cup.m om¢.m mmm.e mmm.~ men * 4<HOh

 

 

 

oo. o.o o.zo m.- ..o. ~.o _.m z ooooz
ooo oo. o.o o.mz o.om o.oo o.o. o.o ooooooz

ooo._ oo. ..o o.oz o.om o._~ o.om o.o zoooooz
o z<o_ooz-zoz
omo.o oo_ o.o ~.oo o.o_ m.o_ o.o. ~.m ozooooz-zoz
oom.~z oo. o.o z.oo o.o. o.o. o.o o.o ozoz
z z .82» o_ .ooz op oz: u o : ozz : o z z_oo oz: o.oso: z_oo ozz o.o: mooozo ztoooooo
o.ooo o.ozo .oooz o.osoz .oooz ozoz .oooz o.oooz oooz o.o: Loooou ozoooo

ozooo z oz: oo ooooooomoz to oozoooooo_oz

 

 

zoom.om u z: Aozzo ozozooooz to ooo: oo ooooooomoz
me owzmceooo_em woo mooezw acemepoo Leoeou ozoocH mo mueeeceomem oe cemoooocumoo Homeceo .ep epeoh

62

Table 15. Percent Distribution of Respondents by Poverty Level Index
and Intake Factor Category Groups (N = 24,362)

 

 

Factor Intake Category Groups %

 

 

 

 

 

 

 

 

TOTAL
NON- NON-MEDICAL
Poverty Level Index NONE MEDICAL 8 MEDICAL MEDICAL % #
0.00 - +0.99 9.1 9.6 11.4 14.4 10.0 2,425
+1.00 - +1.30 6.2 5.7 8.2 8.6 6.2 1,521
+1.31 - +1.50 3.0 3.2 2.4 3.8 3.1 758
Below Poverty
Level: Subtotal 19.0 18.5 22.0 26.7 19.3 4,704
+1.50 - +1.99 9.8 9.0 8.2 10.6 9.4 2,299
+2.00 - +2.99 18.8 19.2 14.9 16.0 18.6 4,541
+3.00 - +3.99 12.3 13.5 12.4 9.5 12.7 3,090
+4.00 - +4.99 6.2 7.1 6.2 5.0 6.5 1,587
+5.00 - +9.99 7.1 9.1 9.4 7.9 8.1 1,968
+10.00 and Above 0.8 1.1 1.3 0.9 0.9 230
Above Poverty
Level: Subtotal 55.1 59.1 52.3 49.9 56.3 13,715
TOTAL Answered 74.1 77.6 74.3 76.6 75.6 18,419
TOTAL Not
Answered 25.9 22.4 25.7 23.4 24.4 5,943
GRAND TOTAL % 100.0 100.0 100.0 100.0 100.0
GRAND TOTAL # 12,308 9,620 1,454 980 24,362

 

.epeopwo>o cog: eemo we: oewuoeoee m.eom; eposmw mg» can» :ewuoeoee
Lew Lezmco e: eon ewes ww use .Aozzv eFegemoez we now: e—oaew eco o.oE seen ems: cewuoeoeu m.eoe: epoz
.1

 

63

 

Nmm.om mm mn~.e omo.¢ mop.m oo~.m omo.m mmo.~ w 4<wow
cop o.o m.- o.op e.mm o.m_ m.m m.w a 4<wow
omm cop m.o P.op m.m_ o.om n.mp N.e_ o.o— o<ofiomz

omo.p oo_ ~.o o.e_ m.mp w.- o.op o.~p m.N_ 4<uHomz a
o<umouzuzoz
omo.m co. F.o m.wp w.n~ m.om o.m_ o.o ~.m 4<0Hamzlzoz
mom.m_ oo_ P.o m.N_ m.ep o.om «.m— o.w ~.m mzoz
. meoecw
w z <z ooze omen emeppeu euooeoco peezem euooeoce peezem accomeou
\eoooeoco weem —ee;em new: Peezem eeozw meow zeoeo: exoocm

eme__eu sow: esem Lessozu ego ecez

 

4<w0h N upezmmee: we now: we pm>m4 :ewueezem

 

 

. “Nom.om u z: o_ozomooz wo oooz
we Fe>e4 :ewpoeoem eco mooecw ocemeuouzooeo: ozoocH muoeeeeomem we :ewuoowcumwo Homecmo .ep epoow

Table 17.

Groups (N = 24,362)

64

Percent Distribution of Respondents by Working Status of
Head(s) of Households (HHD) and Intake Factor Category

 

Intake Factor Category Groups %

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TOTAL
Working(S§atus
of Head 5 of
NON- NON-MEDICAL

House“°‘d$ NONE MEDICAL & MEDICAL MEDICAL % #
MALE HEAD ONLY
Working 1.9 2.9 2.5 2.4 2.4 578
Not Working 0.9 1.4 2.6 2.9 1.3 310
Subtotal 2.8 4.3 5.1 5.3 3.7 888
FEMALE HEAD ONLY ‘7
Working 7.6 9.6 7.9 4.5 8.3 2,016
Not Working 8.3 8.3 18.0 19.5 9.3 2,265
Subtotal 15.9 17.9 25.9 24.0 17.6 4.281
MALE AND FEMALE HHD
Male Only Working 36.6 34.5 28.0 24.9 34.8 8,459
Female Only Working 2.9 2.5 3.7 3.4 2.8 684
Both Working 33.4 34.0 21.2 16.8 32.2 7,844
Subtotal: Working 72.9 71.1 52.9 45.1 69.9 16,987

Neither Working 8.4 6.7 16.0 25.6 8.9 2,155
Male 8 Female HHD 81.3 77.8 68.9 70.8 78.7 19,142
Subtotal Working_ 82.4 83.6 63.4 52.0 80.5 19,581
Subtotal Not Working_17.6 16.4 36.6 48.0 19.5 4,730
TOTAL ANSWERED % 100.0 100.0 100.0 100.0 100.0

# 12, 77 9,607 1,454 975 24,311

(% of Grand Total) (99.7) (99.9) (99.9) (99.5) (99.8)
TOTAL "NO ANSWERS" ( 0.3) ( 0.1) ( 0.1) ( 0.5) ( 0.2)
GRAND TOTAL % 100.0 100.0 100.0 100.0 100.0
GRAND TOTAL # 12,308 9,620 1,454 980 24,362

 

65

Table 18. Percent Distribution of Respondents by Intake Factor Category
Category Groups and Four U.S. Census Regions (N = 24,362)

 

 

Four U.S. Census Regions % TOTAL

 

Intake Factor North- North- No
Category Groups East Central South West Answer #

 

 

NONE 27.6 22.0 20.7 15.3 14.4 12,308
NON-MEDICAL 32.9 28.7 17.1 11.4 9.8 9,620
NON-MEDICAL

8 MEDICAL 48.9 36.9 7.7 3.8 2.7 1,454
MEDICAL 51.1 32.6 9.6 3.9 2.9 980
TOTAL % 31.9 26.0 18.1 12.6 11.4
TOTAL # 7,772 6,330 4,406 3,071 2,783 24,362
CENSUS DATA* % 22.9 27.0 32.4 17.7+

 

* From: Table No. 29, U.S. Bureau of the Census, Statistical Abstract
of the U.S., Washington, DC, 1978 (99th edition). Based on the 1977
Census of U.S. population.

+ Excludes Alaska and Hawaii in that the 1977-78 NFCS only included
the 48 continental states.

66

respondents living in female headed households with fewer than four
persons. These groups were composed primarily of adults living in the
North-East and Central census regions. Both of the groups had more in the
lower education level categories than did the NONE and NON-MEDICAL groups.
The MEDICAL group had the largest percentage of respondents in the not
working status categories and in the below poverty level income group.
The NON-MEDICAL 8 MEDICAL group had a mixture of working and non-working
head(s) of households.

The percent distributions of respondents by urban and rural
urbanization categories also were investigated. All four intake factor
category groups had comparable proportions of urban (approximately 68%)
and rural (approximately 32%) categories.

The findings from the crosstabulations of the factor category groups
with the socio-economic descriptors were compared to the HANES data (NCHS,
1978). From the HANES data the percent of the population 12-74 years on a
special diet were determined by specific age categories and family income.
In the older age categories, more persons reported special problems than
younger age groups (for example, the 12-17 years age groups and the 65-74
years age groups had 3% and 21%, respectively). An inverse relationship
existed between family income and percent of the population identifying a
special diet. Thirteen percent of families with income less than $4,000
and nine percent of families with incomes $15,000 and more reported
special diets (NCHS, 1978).

The cross tabulation of intake factor category groups by socio-
economic variables further supported findings from previous
investigations. Todhunter (1976) found over half of non-
institutionalized elderly living alone. Learner and Kivett (1981)

reviewed the literature on elderly and reported greater longevity'of women

67

and acknowledged widespread chronic disease conditions and poor health.
They (Learner and Kivett, 1981) found low income and educational levels in
their population study of rural elderly in North Carolina. The findings
from our investigation were similar to those of TDdhunter (1976) and
Learner and Kivett (1981) in the descriptions of the two groups self-
reporting medical problems. The two groups, NON-MEDICAL 8 MEDICAL and
MEDICAL, had more older adults living in smaller households with lower
incomes and lower educational levels than the other categories.

Combining the findings from this section with previous sections
expands the typology of each of the groups. For instance, the NON-MEDICAL
category could be characterized as children and adults (25-55 years),
living in households of three or more with working head(s) of household(s)
and incomes above the poverty level. This group was composed of snackers
also eating two or three meals a day. They further reported “not liking
certain foods" and "not liking to eat breakfast early in the morning“ as
affecting the way they ate or drank. Conversely, the MEDICAL group may be
characterized as adults 55 years and older, living in one or two person
female headed households in the North-East or North Central with lower
incomes and educational levels, and larger percentages not working. This
group tended to eat the three meal a day pattern and snack. These
respondents identified themselves as "being (N1 a medically prescribed
diet" or "having a medical problem like diabetes or allergies. "

The information obtained from these typologies can be used for
targeting of nutrition messages or product marketing strategies. For
example, nutrition educators may want to develop nutrition information
for parents of children regarding breakfast or morning snacks for the "on-
the-go working parent.“ Or the nutrition professional developing

information for the elderly population may first want to assess existing

68

medical problems and then gear information for low income and education
levels, noting that this group is apt to be eating three meals a day. ‘The
food industry marketing a new product for the elderly may only want to
make it accessible in the Northeast and North Central regions and may want
to use marketing strategies aimed at low income households at low cost,
for persons with medical problems. These are just a few examples of the

possible uses of the information obtained.

Nutrient Quality Assessment

 

A composite index combining the marginality score (MS) for seven
nutrients and the PFC index (see Chapter III) was used to assess nutrient
quality of intake factor category groups. The MS crosstabulated with the
intake category groups is illustrated in Table 19. This table showed that
the NONE (65%) and the MEDICAL (61%) groups had more respondents with one
or no micronutrient intakes at levels less than or equal to 59.9 percent
of the RDA. Conversely, the NON-MEDICAL (49%) and the NON-MEDICAL 8
MEDICAL (51%) groups had more respondents with two or more micronutrient
intakes at the marginal level of intake.

The ratio of protein, fats and carbohydrates in the diets of
respondents as categorized in the PFC index is shown in Table 20. The
four intake factor groups had comparable proportions in the all "okay"
(approximately 20%) versus all other categories. In all groups, fat was
the most likely to be outside of the specified ranges used in the
development of the PFC index. Both fats and carbohydrates were the second
most likely macronutrients to be outside of their specified ranges in the
ratio. The all "okay" category had the third largest percent of
respondents in each group. All (five) other possible combinations had

less than or equal to seven percent of the total in each group.

69

Table 19. Percent Distribution of Respondents by Intake Factor Category
Groups and Number of Marginal Nutrients In Marginality Index

 

 

 

 

(N = 24,362)
Number of Marginal Nutrients % TOTAL

‘ Four -
Intake Factor None One Two Three Seven #
Category Groups
NONE 47.3 17.8 11.5 8.1 15.7 12,308
NON-MEDICAL 33.9 17.0 13.3 10.8 25.0 9.620
NON-MEDICAL

8 MEDICAL 30.8 18.2 12.7 12.3 26.1 1,454
MEDICAL 39.8 21.6 11.3 10.1 17.1 980
TOTAL % 40.7 17.6 12.3 9.5 19.9 24,362

 

*
Marginal was defined as less than or equal to 59.9 percent of the RDA
for 3 day average intakes. Micronutrients included: calcium, iron,
magnesium, vitamin A, vitamin B6, vitamin 812, and vitamin C.

Table 20. Percent Distribution of Respondents by Intake Factor Category
Groups and PFC Index Categories (N = 24,362)

 

 

PFC Index Categories % TOTAL

 

Intake Factor Fat outside Fat 8 CHO out- All
Category Groups All "okay" of range side of range Others #

 

NONE 19.2 48.6 30.1 2.1 12,308
NON-MEDICAL 20.0 43.5 32.0 4.5 9,620
NON-MEDICAL

8 MEDICAL 22.4 37.2 33.4 7.0 1,454
MEDICAL 22.0 39.9 33.6 4.5 980

TOTAL % 19.8 45.5 31.2 3.5 24,362

 

70

A 12 category (A-L) composite index for assessing nutrient quality
was derived from combining the marginality score and PFC index (see
Appendix A-5 for description of each group). Category A contained those
respondents that had no micronutrient intakes labeled marginal (5;59.9%
of the RDA) and their ratio of PFC was within the specified "okay" ranges.
Categories A, B, C contained those respondents who had no marginal
micronutrient intakes, and were "okay“ (A); one outside of the range (8);
or two or more outside the ranges (C) in their PFC ratio calculation.
Category L contained those respondents who were marginal in three or more
micronutrients and had two or more macronutrients in their PFC index
outside of ranges. The combination of categories J, K, L contained those
respondents with four or more micronutrients at 59.9 percent of the RDA
and their PFC ratio was all "okay" (J); one outside the ranges (K); or two
or more outside the ranges (L). The categories.C, F, I, L contained those
respondents who had their PFC ratio in the two or more macronutrients
outside the ranges category and had none (C); one (F); two (1); three or
more (L) marginal micronutrient intakes.

The percent distribution of respondents by factor intake category
groups and composite index (Table 21) showed the two groups identifying
non-medical factors, NON-MEDICAL 8 MEDICAL (15%) and the NON-MEDICAL
(14%) with the most respondents in the L category. When the composite
groups reflected the PFC index, the intake factor groups were more alike
than when they reflected the marginality score. For example, C, F, I, L
were composite categories with two or more macronutrients outside of
specified ranges (see above explanation). The NONE, NON-MEDICAL, NON-
MEDICAL 8 MEDICAL and MEDICAL included a total of 31, 35, 38, and 37
percents of the composition C, F, I and L index categories, respectively.

For the categories where three to seven micronutrients were marginal (J,

71

K, L), NONE (24%), NON-MEDICAL (36%), NON-MEDICAL 8 MEDICAL (38%), and
MEDICAL (27%) were more different.

Table 22 showed that females had the larger percentages in the J, K,
L (38%) or 1. alone (14%) composite index categories and the smaller
percentages in the A, B, C categories (32%) compared to males (53%).
Blacks had slightly more in the lower (J, K, L) categories and fewer in
the higher (A, B, C) categories than whites. The crosstabulations of the
composite index with RDA sex/age categories showed that the younger age
categories regardless of sex had larger percentages of respondents in the
A, B, C categories. The older age/sex groups had more in the J, K, L
categories than the younger age/sex categories.

Snacking behavior by six representative composite index categories
showed more snackers in the A, B, C categories than non-snackers (Table
23). Table 23 also illustrates that the 2,2,2's had the fewest (17%) in
the A, B, C categories and the most (54%) in the J, K, L categories than
the other major meal pattern groups. There were descending directional
differences with the 3,3,3's having the most and the 2,2,2's having the
least in the A, B, C categories. An inverse ascending relationship showed
for the J, K, L categories from the 2,2,2's to the 3,3,3's. Educational
levels demonstrated a similar trend as snacking behaviors (Table 24). The
higher the educational level attained by the head of household the larger
the A, B, C category percentages and the fewer in the J, K, L categories
(and vice versa). These tables indicate that meal and snacking patterns
and educational level may have effects on nutrient quality intake. These
could be significant factors (regardless of intake factor category group)
that may ultimately qualitatively differentiate food consumption
patterns. More careful data collection of data on snacking behavior is

needed to quantify the differences.

72

 

 

 

 

 

 

 

 

 

Noe.wm m.o_ e._w P.o o.o o.o N.~ o.o w.m ~.m m.ww o.- o.o wows:
mo~.m o.m_ m.mw o.o o.o o.o o.o m.m m.o o.o o.o m.n_ o.m xeowm
oom.o_ o.o_ o.o_ w.o o.o o.o o.o o.o ..o o.o o.w o.~p o.o ozoao:
ooo.o_ m.~ A.“ ..o w.o o.o o.o ..N A.“ ..o o.o, o.oo o.o opoz
w o x o H I o w m, o u m < moo:
eco xom
o<wow a mmwzemeuou xeecw euwmeoeeu

mom.om u zv mewzemepou xoeoH epwmeoeeu oco eeom oco xom on moooeoeomez we :ewuoowcumwo oceezeo .NN ewoow
Nom.o~ o.o— o.__ o.o o.o m.m o.~ o.o o.n m.m m.wp w._~ o.o & o<wow
0mm ~.ow N.o_ o.o o.o m.m N.N m.w o.o o.o N.NP m.mw o.o 4<uwomz

omo._ w.m_ m.~_ o.o_ o.o o.o m.~ N.~ o.o N.m m.o_ o.ow m.m o<u~omz o
o<u~omzizoz
omo.o o.o, m.ow o.o o.o o.m m.~ o.o o.“ —.m o.o m.~w m.o o<u~omzuzoz
wom.m_ o.o _.ow _.m o.o _.m m.~ m.e _.m o.m o.~— _.o~ o.m mzoz
* 4 x a ~ - x w w m o u m < mazoww Awommpou
o<wow a mewcemmooo xee:~ euwmeoseu Lepeow exoucH

 

xeeew euwmeoaeu eco mooezo aceoeoou Leoeow ozoocw mm moomeoeomem

zooo.oo " zo mowzoooooo

we :ewooewgomwo oceeceo

.—N mweoh

73

.eeeopexm ow :Lmuuoo

poms =zezue..esu memeoweow ego mczeuuoo woes LenoE e ecu A—oe eoo mewzemeuou xeecw euwmeosee
e>Pezu one we xwm mowoooee awee epooo esp pogo aeow ego ea woe Nom.o~ ecu coco mmow ow 2 wow A

 

 

 

 

 

 

~.N~ o.mp m.- m.m P.w m._ N.~.N

P.ww ~.o~ w.o_ m.w N.ww e.m N.~.m

m.NF o.ep o.o N.ww m.op o.o «.m.m

o.o o.m m.o w.m_ ~.wm w.m m.m.m

mczepuoo woe:

pcmoeezw owe: Loom

m.m— m.mw a.“ P.pp o.mw o.o meexeoemucez

m.m o.ww N.m m.ww w.m~ o.o mzexeocm

Lew>ozmm mewxeocm

o.ow m.F— o.o m.—w w.wm m.w o<wow
o x a o m <

 

& mewcemeuou xoeom mowmooeeu

 

 

mmwzemwuoo xeenH euwmeoseu m>wuouemmecoem xwm eco *Amom.mfi u xv mcceuuoo woe: ucmoeeww
owe: coed eco onowo.wH u 2V Lew>ogem mcwzeocm we muceeoeomem we mcewuoowcumwo aceeceo .mm eweow

74

 

 

 

Nom.o~ o.ooH o.o“ o.o” o.o o.o m.m o.~ o.o o.o m.m m.- e.- m.» u o<wow
m-.o m.w~ o.o o.o o.o o.o ~.m w.~ o.o m.~ m.m w.~— o.om o.o .eoLu omen eeo
.eoLw emu—pee
cmo.o o.eH m.m e._w o.m m.o ~.m m._ m.o o.o m.~ m.~— m.e~ m.~ eme__eu
«sew
mo~.m o.om o.__ m.~_ _.o o.o o.o m.~ o.o ..m w.~ p.p— m.- o.~ .eoLw
_eezem new:
coo.m «.mH m._w o.~w m.~ o.o N.m N.~ m.m o.o m.m m.ww o.om w.~ .eezem
:3: 2.8
o~o.~ m.m ~.op w.ew o.o o.m ~.m o.~ m.~ o.o ~.o ..o— o.~_ ~.o .eozu
_ee;em Lossozc
mmo.~ m.m m.e_ o.mw w.o_ o.o o.o o.o m.m m.o o.o o.o o.m_ ~.o peezem eeozo
eeem eoo eoez
w a

o x o _ z o w m o u m < mpe>eo oewuouoeu

o<wOw a mewzeoeuou xeeo~ mowmeoseu

 

 

son: . z: 3:333 .32; 33858 o... 223 5:833 a 3523.2. to 5:323; 28:... .oo 22:

75

Tables with crosstabulated specific marginality micronutrients
against marginality score categories showed: vitamin 86 (37%) and calcium
(32%) with the most respondents in the 5 59.9 percent category and
vitamin B12 (10%) with the least. The two intake categories with non-
medical factors included, NON-MEDICAL and NON-MEDICAL 8 MEDICAL,
consistently had larger percentages of respondents in the 5.59.9 percent
RDA nutrient groups than the NONE and MEDICAL groups. These findings were
consistent with findings of Friend and Marston (1975) on the availability
of micronutrients in the U.S. food supply in RDA percentages for male and
female adults. Vitamin 86’ calcium and magnesium were least available and
vitamin B12 was most available.

The large percentages of respondents which had either fat outside the
specified ranges in the PFC index or fat and carbohydrate outside of the
ranges was an expected finding. Page and Friend (1978) discussed the
changes in percentage intake of calories from energy yielding nutrients
from 1909-13 to 1976 (based on per capita civilian consumption). In the
earlier years the ratio was protein (12%)-fat (32%)-carbohydrate (56%).
In 1976 the ratio changed to protein (12%)-fat (42%)-carbohydrate (46%).

Nutrition Education Applications

The combination of information from these series of nutrient quality
assessment analyses with previous sections of this analyses expands the
typologies of factor intake category groups. Although the NONE and the
NON-MEDICAL categories were more alike in socio-economic descriptors, it
was the MEDICAL and the NONE groups that were more alike in meal and snack
patterns and subsequent nutrient quality assessment. The NONE and the
MEDICAL groups had more respondents in the 3,3,3 pattern and it was these

groups which had greater than 60 percent of these respondents in the one

76

and none categories for marginal nutrient intakes. So, although the two
groups reporting non-medical factors had the most respondents snacking,
they still had more respondents with marginal nutrient intakes. Crocetti
and Guthrie (1982) found the 2,2,2 pattern to have the most individuals
with “less desirable“ nutrient quality intakes. They also found the
2,2,2's to have the most snackers as compared to the other four major meal
patterns.

The nutrition education and marketing strategy implications are
numerous. The following is a specific example of how this information may
be used. In targeting a nutrition education effort or a product
advertisement for households with children and adults (25-55 years),
professionals will want to acknowledge that this group will be mixed with
three meal a day eaters and two and/or three meal pattern eaters. In
addition, this group may have large percentages of snackers. Because of
the large percentages that identified "not liking to eat breakfast early
in the morning," a mid-morning snack may be promoted. Having identified
that calcium and 86 were the micronutrients most likely to be "marginally"
ingested, a mid-morning breakfast/snack for persons who don't like to eat
as soon as they get up may be appropriate. Perhaps the snack could be
whole grain cereal and milk. Stress given to low fat milks to incorporate
the understanding that fat was high in the PFC ratio. From Crocetti and
Guthrie (1982) findings it was determined that fat was outside of the
specified ranges on the high end of the range and carbohydrate on the low

end of the range (see Appendix A-2).

77

Food Group Intake and Personal Behaviors

This final set of descriptive variables was included to define
specific food groups which respondents were eating. Additionally, the
eating alone and eating out behaviors were described, as well as food
supplement use and weight status (see Chapter II for descriptions of
variables). The information obtained from these analyses supplemented
findings from previous sections.

The percent distributions of respondents by "total number of
mentions" for the Crocetti food groups are presented in Table 25 (food
group 32: nutrient supplements, meal extenders and "health foods“ was
excluded because less than one percent of the respondents even mentioned
this group in the three days). The 32 food groups were developed by
Crocetti (1982) to reflect use and function of food items in the 0.5.
diet rather than nutrient content. Table 25 shows a small percent
( 5_25%) of the total population mentioned the following food groups at
least once: 'fruitades; alcohol; poultry Inixtures; fish, shellfish,
seafood and mixtures; legumes; spoon desserts; candies; and nut butters.
This was true for all four of the intake factor category groups. More
than 75 percent of the respondents mentioned the following groups at
least once: nfilk, meats, starchy side dishes, vegetables, bread, and
garnishes. This also was consistent across the intake factor category
groups.

The milk group was differentiated further to show total number of
mentions by intake category groups (Table 26). This table showed that the
NONE (13%) and MEDICAL (14%) had the fewest respondents in the no mention
category and the most in each of the categories three-to-five and more.
The persons reporting no problems and only'medical problems mentioned more

fluid milk items in their three day intake records. This supplements

78

Table 25. Percent Distribution of Respondents by 31 Crocetti Food Groups
and Total Number of Mentions of Food Groups (N =24,362)*

 

 

Total Number of Mentions of Food Groups %

 

31 Crocetti

 

Food Groups 0 1-2 3-5 6 or more
Milk 16 19 32 33
Frt/Veg Juice 55 25 19 l
Coffee/Tea 34 14 24 28
Soft Drinks 41 31 22 6
Fruitades 86 ll 3 +
Alcohol 87 8 1
Soup 70 27 +
Eggs 47 43 11 +
Cheese 56 37 8 +
Meats 5 32 51 11
Meat Mixtures 58 38 4 +
Poultry 55 4O 4 +
Poultry Mixtures 94 6 + +
Fish/Seafood 75 24 1 +
Fish/Seafood Mixtures 91 9 + +
Legumes 78 20 l +
Starchy Side/Main Dishes 14 53 31 2
Vegetables 8 29 41 23
Starch/Veg/Protein Mixes 63 35 2 +
Bread 3 16 49 33
Cereal 48 34 17 1
Fruits 42 33 20 5
Frozen Desserts 69 26 5 +
Spoon Desserts 82 17 l +
Cakes, Cookies, Pies 34 38 23 5
Snacky Foods 53 36 10 l
Candies 85 12 2 +
Garnishes 24 4O 28 9
Butter 8 Margarine 32 39 25 4
Sugar Garnishes 33 31 25 11
Nut Butters 82 16 2 +

 

* N = 24,362 for each of the 31 Food Groups.
+ Less than 1 percent.

79

Table 26. Percent Distribution of Respondents by Intake Factor Category
Groups and Total Number of Mentions of Fluid Milk (N = 24,362)

 

 

Total Number of Mentions of Fluid TOTAL

 

 

,.
Intake Factor _1n1k %
Category Groups 0 1'2 3‘5 5'8 9 #
NONE 12.6 17.4 33.4 24.0 12.8 12,308
NON-MEDICAL 18.6 20.9 31.2 19.2 10.1 9,620
NON-MEDICAL

8 MEDICAL 21.9 21.9 28.2 16.8 11.3 1,454
MEDICAL 13.9 19.6 33.2 19.9 13.5 980

TOTAL % 15.6 19.1 32.1 21.5 11.6 24,362

 

80

findings from the nutrient quality assessment analyses. These categories
had the fewest respondents in the multiple marginal nutrient intake
categories. Calcium was the micronutrient with the largest percentages of
respondents with f_59.9 percent RDA intake. Milk is an excellent source
of calcium.

Other food group consumption trends were obtained from the cross
tabulation of intake factor category groups by total number of mentions in
the three days recorded for specific food groups. The two categories with
medical problems seemed to be more alike in their consumption of foods and
the NON-MEDICAL and NONE categories appeared more similar. For example,
the categories which reported medical factors, NON-MEDICAL 8 MEDICAL and
MEDICAL had more respondents mention: more fruit and vegetable juice
items; more coffee and tea; fewer soft drinks; fewer
starch/vegetable/protein Inixture combinations; more fruit items; and
fewer cakes, cookies, pies, and pastry items. For the aforementioned food
groups the opposite held true for the NONE and NON-MEDICAL categories.
This may be related to the similarities in socio-economic composition of
the categories.

In Tables 27-30 are shown the percent distribution of respondents by
intake factor category groups and eating out and eating alone behavior use
of vitamin and mineral supplements, and weight status. Tables 27 and 28
illustrate that the respondents who reported some type of medical factor
were more similar and the NONE and NON-MEDICAL groups were more alike. The
persons reporting medical factors recorded eating alone and eating at home
more often. This was expected given the results of the socio-economic
descriptor analyses. These groups had the most older respondents living
in smaller households with lower incomes. The NONE and NON-MEDICAL had

the most working adults in larger households with children and teens.

81

Table 27. Percent Distribution of Respondents by Eating Out Behavior*
and Intake Factor Category Groups (N = 24,362)

 

 

Intake Factor Category Groups % TOTAL
NDN- NON-MEDICAL
Eating Out Behavior NONE MEDICAL & MEDICAL MEDICAL #

 

All Meals At Home

 

 

 

 

 

 

All Snacks at Home 10.9 11.1 17.3 19.5 2,858
Some Snacks Out 3.4 4.1 2.8 2.4 878
All Snacks Out 0.1 0.2 0.1 + 28
No Snacks Reported 11.9 9.5 15.7 18.3 2,785
Incomplete Snacks 13.7 12.6 18.6 18.3 3,345
SUBTOTAL 40.0 37.5 54.5 58.5 9,894
Some Meals At Home

All Snacks At Home 10.9 9.9 8.7 7.6 2,496
Some Snacks Out 11.5 17.0 10.1 8.2 3,275
All Snacks Out 0.6 0.9 0.4 0.1 163
No Snacks Reported 13.7 9.6 .7 8.6 2,786
Incomplete Snacks 22.6 23.5 18.3 16.2 5,467
SUBTOTAL 59.2 60.9 44.2 40.6 14,187
All Meals Out 0.4 0.8 0.4 0.4 143
Incomplete Meals 0.4 0.8 0.8 0.5 137
No Answers + + .1 + 1
TOTAL % 100.0 100.0 100.0 100.0 24,362

 

* Eating out behavior pertains to food eaten away from home and not
from household food supply, (e.g. a brown bag lunch or picnic packed
from home food supply would not be considered as having been eaten out.
+ Less than 0.1 percent.

82

Table 28. Percent Distribution of Respondents by Intake Factor Category
Groups and Three Day Patterns of Eating Alone (N = 24,362)

 

 

 

 

3-Day Patterns of Eating Alone % TOTAL

Intake Factor All Meals Some Meals No Meals No
Category Groups Alone Alone Alone Answer #
NONE 2.4 64.8 31.9 0.9 12,308
NON-MEDICAL 4.0 73.6 21.6 0.7 9,620
NON-MEDICAL

8 MEDICAL 12.1 71.0 16.1 0.8 1,454
MEDICAL 10.7 65.6 22.1 1.5 980
TOTAL % 3.9 68.7 26.5 0.8 24,362

 

Table 29. Percent Distribution of Respondents by Intake Factor Category
Groups and Reported Use of Vitamin and Mineral Supplements
(N = 24,352)

 

Reported Use of Vitamin and HineralSupplements% TOTAL

 

 

N0 and Total: Yes, Yes, Yes, NA
No Answer YES Regular Irregular t0 Reguo #
1arity*
NONE 71.1 28.9 20.3 7.2 1.4 12,308
NON-MEDICAL 64.4 35.6 23.4 10.0 2.2 9,620
NON-MEDICAL
8 MEDICAL 55.7 44.3 33.6 8.3 2.4 1,454
MEDICAL 60.0 40.0 30.5 6.6 2.9 980
TOTAL % 67.1 32.9 22.7 8.4 1.8 24,362

 

* Indicated that the respondent did not indicate regularity of use, but
did identify kind of vitamin or mineral supplement.

83

Table 30. Percent Distribution of Respondents by Intake Factor
Category Groups and Weight Status (N = 24,362)

 

 

 

 

Weight Status % TOTAL

Intake Factor ‘Under Over
Category Groups Weight Normal Weight NA* #
NONE 5.0 72.9 21.3 0.9 12,308
NON-MEDICAL 4.4 71.2 23.3 1.1 9,620
NON-MEDICAL

8 MEDICAL 3.8 57.8 37.4 0.9 1,454
MEDICAL 3.0 67.6 28.5 1.0 980
TOTAL % 4.6 71.1 23.3 1.0 22,362

 

* NA = percentage of respondents that had no answers to either height
and/or weight and those adults where their height was too
extreme and weight status could not be calculated based on
charts provided by Food and Nutrition Board (1980).

84

Therefore, it was expected to find these categories having larger
distributions in the non-eating alone categories and eating out
categories.

Due to questionnaire wording it is difficult to interpret the reported
use of supplements. However, from Table 29 the NONE category had the most
(71%) in the no usage and no answer category. The NON-MEDICAL 8 MEDICAL
category had the most respondents (44%) in the reported total usage of
supplements category.

The NON-MEDICAL 8 MEDICAL had the most respondents (37%) in the
"overweight“ weight status category. The NONE group had the most (73%) in
the "normal“ weight status category.

From the analyses of food intake and personal behaviors several trends
were seen. For example, milk, meats, starchy side and main dishes,
vegetables, bread, and condiments or garnishes were used by large
percentages of the study population regardless of whether or not they
reported a problem or factor affecting their intake. Approximately 99
percent of each of the factor category groups were eating either all meals
at home or some meals at home. More than 21 percent of each of the
categories fell into the "overweight" weight status category. These
findings were consistent with trends in U.S. dietary patterns described by
Sanjur (1982) and Pao (1981). Sanjur (1982) summarized findings from
various national surveys. The following U.S. consumption trend were
reported by Sanjur (1982): increases in milk and meats, decreases in
fruits, decreases in potatoes, increases in dark green and .yellow
vegetables, increases in oils, sugar and other sweetners, and decreases in
butter, coffee and tea. Pao (1981) derived estimates of percentages of
consumption by NFCS respondents for specific food groups based on nutrient

content of foods. The findings from this investigation of persons with

85

self-reported factors affecting their intake were within the ranges for
the food groups as Pao (1981) analysed them. Pao's analysis provides a
check on the findings of this investigation.

Obesity has increasingly been characterized as a major American
health problem. Its prevalance (as much as a third of our population has
been reported as being obese), its relation to other medical conditions,
and its resistance to treatment have been the reasons for its recognition
(Concern, 1979). Although NFCS self-reported weight and height were used
to calculate weight status, other investigators have documented accuracy
or underreporting in these self-reported measures. Stunkard and Album
(1981) found Americans were remarkably accurate in their self-reporting
of weights. Pirie et a1. (1981) found women's self-reported weight to be
frequently low and an under-estimation was found among heavier men. In
either case, the large percentages determined to be in the "overweight"
categories would not be diminished substantially. Therefore, the NFCS is
additional documentation of the prevalence of "overweight" persons in our
society based on recommended ranges from the Food and Nutrition Board

(NAS, 1980).

Composite Typologies of Intake Factor Category Grogps

In Phase I of this investigation, factor intake category groups based
on self-reported non-medical and medical factors or problems potentially
affecting dietary intakes of NFCS respondents were derived. These groups
were analysed in terms of: meal and snack patterns; socio-economic
characteristics; nutrient quality assessment; and food intake and
personal behaviors. The composite discriminating descriptive
characteristics of the intake factor groups is presented in Figure 3.

The NONE and NON-MEDICAL groups included the largest percentages of

863

 

meoegm Lozoe coca
eeuwe aces wee;
oo eoo ecewo uom
meocog eewmaoe
ewuoc uwo saw: was
=POCPGLME:
mooewcaoeezews
H Le e: new: age

Azkoo o.o.o
wocuoeu

eeo umomsugez cw e>w4
mpezep

—ocewoouoee ooze;
o.o;emoe; we come

mewxzezlee: eeo o.osew
eseecw

Pe>ep auge>eo :e_em

NIH ee~wm epezemoez

foe» mm.» 3 33

.xmLepwo co
Le meoeoowe exw_
Empeezo poeweee o e>oz
.oewe
eeowzumezo Leoeee oo

 

4<u_ouz

eswo on» we eEem
ecepo ego use; we won
..HnmwOEm>O:
memouoeeceo omemzoo
oncogewooom em:
memoog eewmuoe
ewuog one saw: was
=FUCFGLME:
muoewcoooecews
aces Le N cow: new
Azooo o.o.o
-ceo eoo Annoy o.o.m

—oLu=eu

ego umomzucez cw e>wo
ewesemoe;
we Amveoe; mewxcez

-oee eco ocwzcez eoxwz
m_e>e_

pooewuoeoee Lezeo

m-“ eerm e—egemoe:

ozooz o~-oo oo_oo<
.uewe
emowzemezo Leoeee
o co ego Le\eoo

Eepoezo wouwees o e>ox
.meeew

:wouzee ezw_ o.oee
.azmwez

emep e» uewe o co
.Eego sow:

eewuo o.oee meeew usem

 

4<u~omz a 4<u~omzuzoz

ecepo

:euoe uec mpoe:
wee: Eecw mozo

eeo uo :eooe o.oe:
memooc eewmooe

ewuoc one cow: New
=FQCWOLQE:
oueewzoooezewe

ages so m cow: aom
zoooo o.o.o

eco Augmv m.m.m-=ez

o.o;emoe;

we Amoeoez mewzzez
meeeecw

—e>ep hocezeo e>eo<
cote—woe eco

meme» .mu—oeoIA—wsow

meow» mmumm

mopoeo eco eecepwgu

.mcwcges
cw x_Loe umowxoeco
oowooo ozo— o.ooo
.meeew
cwougeu ezwp o.oeo

 

4<u_omzlzoz

eee_o
:euoe we: mpooz
eEe; secw xoxo
ego oo :euoe mpoex
memcoc eewmuoe
ewuoz one new: Ram
o—oewmcoeo
weewcuooecuwe
H Le e: saw: umo

zoooo o.o.o

ewezemoeg
we Amveoeg mowxcez
coco—wee eco
meeeu .mu—oeolxpwsow
mm erm epezemoez
mLoe» m~ ooze—wzu

ecez

 

mzoz

mzew>ogem eeuopem eeew

exoue_ «cowguzz
megeuaoo woe:

muwgoocmeseo

so_ooao
Lo Looooz oooowooooo

 

meawmwceueococu
e>wuewgemeo
mewuoewewcemwo

 

 

mooezu xceoeoou Leueow exoo:_ we mewomwceuuogozu e>wuowcumeo mewuocwswcumwo .m mason;

87

respondents less than 18 years living in larger families composed of
working adults, children, and teens. The actual specific food group
mentions for these two groups were also quite similar. The personal food
behaviors such as eating alone and eating out behavior also were more
alike for these two groups than the groups identifying some type of
medical problem.

The two groups identifying some type of medical problem, NON-MEDICAL 8
MEDICAL and MEDICAL, were comparable in socio-economic descriptors and
specific food group total mentions. These groups had the most:
respondents 55 years and older; smaller households; families composed of
non-working adults; female headed households; and lower education levels
of head of household. These two groups were more similar in intake of
specific food groups and eating related behaviors.

The meal and snack patterns were more alike for the NONE and MEDICAL
groups and for the two groups identifying some type of non-medical factor
affecting intake. This also was true of the nutrient quality assessment
indices. All groups showed similar trends in the diet's ratio of protein,
fats and carbohydrates. Fat was the macronutrient most apt to be outside

the specified ranges in the PFC ratio.

Nutrition Education Implications

The findings from these analyses can be used to target nutrition
education messages or strategies in terms of the demographic
characteristics, meal and snack patterns, nutrient inadequacies, food
eaten or not eaten, and eating behaviors of the study population. This
information may aid in nutrition education or message targeting based on
iNdividuals' actual reported perceptions of factors or problems affecting

intake or special diets. Communication specialists and nutrition

88

educators concur that messages presented in terms familiar and meaningful
to target populations will have more impact than terms which have meaning
only to the educator/communicator.

The descriptive typologies derived (see Figure 3) to characterize
each of the intake factor category groups provide general trends in terms
of demographics, food related behaviors, and nutritional problems to aid
nutrition educators. For example, a group may be interested in promoting
products or developing programs for persons on diets to lose weight. From
the typologies derived from the NFCS data, the groups which may be most
responsive are women 25-55 years of age. In addition, younger aged
females may be a responsive audience. These individuals would likely be:
living in households of three or more persons including children, teens
and adults; have above poverty level incomes; high school or higher
education levels of the head of household; and working male or both female
and male heads of household. Large percentages will be eating the three
meal a day pattern and snacking. Large percentages will also be
inconsistently eating two or three meals a day and snacking. These
persons may not like eating breakfast in the early morning, not like
certain foods or find that some foods may not agree with them. They will
likely ingest greater than 35 percent of their calories from fat and may
be marginal in percent RDA intakes of micronutrients, specifically
calcium, vitamin B6 and magnesium. Theylnay eat some meals alone and some
meals away from home.

Considering the above typologies nutrition educators or planners may
want to promote nutrient dense meal items and/or snacks which have
significant levels of calcium, 86 and/or magnesiunu Foods may be promoted
which can be eaten at home, at work or "on-the-go.“ Items to "break-the-

fast" from the night until “a little later in the morning than just after

89

getting up" may' catch the eye and interest of target populations.
Additionally, planners may want to stress that each individual needs to
recognize his or her own dietary intake pattern and that no particular
foods must specifically be eaten to ensure adequacy. Working around
personal dietary patterns or specific food likes and dislikes need not
affect the quality of food intake. Stressing these points may decrease
target population perception that not eating certain foods or having
particular foods that do not agree with them will negatively affect their
intakes. The typologies determined in Phase I may provide educators
and/or planners with descriptors of specific population segments to be

used in information dissemination, teaching, or marketing products.

90

FOOTNOTE

1Contingency coefficient (C) is calculated by taking the square root of
chi-square divided by the chi-square plus N. When chi-square (x2) is
significant, the C is significant. Both measures of association take
the N of the sample into consideration. Dividing through by the large

sample size (N), all of the x zswere significant.

CHAPTER V
RESULTS AND DISCUSSION: PHASE II

In reviewing the current empirical evidence, no
single factor appears to be responsible for the
development of dietary patterns . Rather, a
constellation of factors ... behaving in a
synergistic fashion appear to be more significant
than any single factor working independently.
Needless to note, we must continue efforts to unravel
these interactive mechanisms, which may prove more
important than the sum of individual determinants
(Sanjur, 1981:xiii)

In an effort to continue to "unravel" the factors which
simultaneously interact in food related behaviors, Phase II of this study
was conducted. The model illustrated in Chapter II was estimated and the
results are given in this chapter. Data obtained in the Nationwide Food
Consumption Survey (NFCS) provided a unique opportunity to apply
multivariate analysis techniques to organize further understandings of
food behaviors. The large sample size and availability of respondent data
on demographics, food intake (and calculated nutrient intake) and related
behaviors made the NFCS a viable data base for this investigation. The
first task in Phase II was to determine the study population and specific
variables to represent the four variable sets in the model. The dependent
variable, problem versus no problem, and the three independent variable
sets (1. demographic characteristics; 2. respondents' nutrient intakes;
and 3. food and personal related behaviors) are statistically described.
Factor analysis, discriminant analysis and canonical correlation analysis

were used to estimate relationships between the four variable sets. The

91

92

results are presented and discussed in the following sections.

When appropriate chi-squares and F-ratios and significance levels
are reported as statistical tests to assess systematic relationships
existing between categories. As in Phase I, due to the large sample size
and the incorporation of N in the tests of association, these calculations
were given minimal consideration by the researchers in discussing results

and determining implications.

Sample and Variable Descriptions

The sample selected for use in Phase I (see Chapter IV) was also used
as the study population for Phase II (N = 24,362). For selection and
description of variables to be incorporated in the model, a 10 percent
random subsample was generated from the study population. The 10 percent
subsample was used to check for assumptions of linearity among the
variables. In Chapter III descriptions are given of the methodology used
to transform and/or select variables which met. with assumptions of
multivariate analysis techniques used to estimate the model.

The construction and description of the dependent variable, problem
group and the three sets of independent variables are given in Appendix
A-4. The dependent variable was a dichotomous variable called "problem
versus no problem". Fifty-one percent of the sample identified some type
of factor affecting the way they ate or drank and/or being on a special
diet. (See page 1 for the specific NFCS questions used to identify
problem variable.)

Descriptive statistics on the total 10 percent random subsample for
each of the indicators of the three sets of independent variables are
given in Tables 31 and 32-34. Table 31 lists the means, standard

deviations, skewness, and minimum and maximum scores for each of the

93

.mewoowco> mooseuezewe eu ewnoewwooo we: mucesezomoez w
.o-< xwecmoo< cw ompmww ego mLeooewecw ego we :eoe Lew mwomooe we mpwoo ezw +
.Ammm.o~uzv H among Eezw cowoowooeo xeoum wove» one we eonomoom seecoz zofi u z

.1

 

 

 

o.o o.ooo o.o o.o o.oo o.ooo owoooza zoz z mzx
o.o o.oo o.o o.o- o.o o.oo ozo mowcowoo z sz
o.o o.oz o.o o.o- o.o o.oo oou mowco_oo z fizz
w o.o o.o z o.o o.o ooocooom oox
w o.o o.o w o.o o.o zowo ~ozoooo ooz
o.o o.oo o.o o.o o.o o.o wo>oz zoco>oo oox
o.o o.oo ooo o.o- o.o o.oo ozoooz ooooooooo mox
o.o o.oz o.o o.o- o.o o.mH ,o_oz oowoooooo oox
w o.o o.o w o.o o.o moo: oox
w o.o o.o w o.o o.o xom No:
o.o o.oo o.o A.o o.oo o.oo moo Ho:
eecowco> eceem 3ezm :ewuow>eo +:oez Leuoewecfi
cwz\xoz Eoewxoz\soewcwz ecoecoum

 

 

.zooo.muzv moom ozoowtoz ooooooooooo ozo wo ozoooowooo w_< to: mowoowooom ozwoowtoooo .Hm ozoow

94

.vld. xwflcmaa< :_. Umumw— mwm mwoumuwflcm 2...... $0 50mm Low mxsmmmE .....o mywca m£H+

 

 

 

o.o o.oo o.o o.o o.o o.o zoo: moowoooz w max
o.o o.o o.o o.o o.o o.o moooom ozowoz oux
o.o o.ooo o.o o.o ~.NN o.oo ozoooo zozooo oooooeoooo: z oozz
o.o o.ooo o.o o.o o.ooo o.ooo o owooowz zoo o mozx
o.o o.o o.o o.o- o.o o.o moo owooowz «oz z ooz Nozz
o.o o.ooo o.o o.o o.om o.o“ oo oosoowz zoz z Hozz
o.o o.ooo o.o o.o o.oo o.ooo owoopwoowz zoo z oozx
o.o o.ooo o.o o.o o.oo o.ooo owoowzw :oz z ozx
o.o o.o o.o o.o- N.o o.o z owooowz :oz z ooo ozz
o.o o.ooo o.o o.o o.oo o.omo moocozooozo ooz z Nzx
o.o o.oom o.o o.o o.oo o.oo oowoooooz zoo z ozx
o.o o.ooo o.o o.o o.oo o.ooo ooLH zoo z ozz
o.o o.ooo o.o o.o o.oo o.oo sowopoo ooz z ozz

wecowco> eceem 3ezm :ewuow>eo com: Leuoewecw

cwz\xoz Eoswxoz\EoEwcwz ecoecoom +

 

 

Au.:eov .Hm mom<o

95

.o-< xweceoo< cw eoomww ego maeuoewecw esp we zoom Lew eaomoee we wows: on»

.mewoowao> mooseuesewe eo ewnoewwooo we: moeesmaomoez w

 

 

+
. . . . z - oxa
x o H o o a o o o o N oo oooxz ooz x
o.o o.o o.o o.o o.o o.o aooooo ooz moowoooz x ooax
o.o o.oo o.o o.o o.o o.o oowaooaoz o aooooo moowoooz x NHax
o.o o.oo o.o o.o o.o o.o moaa .ooxoo moowoooz x HHax
o.o o.o o.o o.o o.o o.o moaommoo ooNoaa moowoooz x ooax
o.o o.oo o.o o.o o.o o.o _ooaoo moowoooz x oax
o.o o.oo o.o o.o o.o o.o oooao ooowoooz x oax
o.o o.ox o.o o.o o.o a.o moxooooooz moowoooz x xax
o.o o.o o.o o.o o.o o.o oozmwo zoaoom moowoooz x oax
o.o o.o o.o o.o x.o o.o moooooo mooaoooz w max
o.o o.ox o.o o.o o.o o.o poo: mooooooz w oax
o.o o.oo o.o o.o o.o o.o oowxooawoo moowoooz z max
eecowao> eceem cowoow>eo
xozxooz soowxoszoswowz zozm oaooooom zoo: aoooowooa

 

 

Au.:ouv .Hm m4m<w

96

.o-< xwecmoo< cw emomww mao maeooowecw moo we zoom Low maomome we wows: mow +

 

 

o.o o.o o.o o.o- o.o o.o aoooom mo: mzoo x axax
o.o o.o o.o o.o- o.o o.o :oooo oo: ozoo x ooax
o.o o.o o.o o.o- o.o o.o omowxooao mo: ozoo x m:x
mooowao> maeom Zmzm cowoow>mo +comz Leooowecw
:w2\xoz Eoewcwz\saawxo: etoecoom

 

 

Au.:eov .fim mom<b

97

TABLE 32. Group Means and Tests for Group Differences For Indicators of

Demographic Characteristics by the Dependent Variable,
Problem Group (N=1,809)*

 

 

 

 

Problem Group i T

Demographic No

Indicators Problem Problem F Ratio p
xm Age 34.8 28.1 56.1 .01 7‘
xD2 Sex 1.6 1.4 18.3 .0176
XD3 Race 1.1 1.1 44.3 .37
XD4 Education Male 13.2 12.9 2.9 .08
XD5 Education Female 12.2 12.0 1.7 .18
XD6 Poverty Level 2.8 2.8 2.5 11
XD7 City C l - Central City 0.2 0.3 53.3 .10
X08 City C 2 - Suburban 0.4 0.3 47.8 .08

 

* The N reflects the sample which had no missing values for each of the
indicators of demographic characteristics of the 10% random
subsample generated from the total study population (N=2406).

t The units of measure for each of the indicator variables are listed
in Appendix A-4.
f p §_.01 .

98

TABLE 33. Group Means and Tests for Group Differences For Indicators of
Nutrient Quality By the Dependent Variable, Problem Groups

 

 

 

 

(N=2,256)*
Problem Groups“ 3 +

Demographic No

Indicators Problem Problem F Ratio p
XNl % Calories Fat 39.6 39.8 0.4 0.5
xN2 % Calories CHO 43.3 43.9 3.0 0.1
xN3 % RDA Protein 156.6 176.7 66.4 0.0176
xN4 % RDA Calcium 82.4 92.1 27.0 0.01f
xN5 % RDA Iron 95.9 108.6 34.9 0.01*
xN6 % RDA Magnesium 77.8 84.2 27.0 0.01?
xN7 % RDA Phosphorous 129.3 139.8 22.2 0.0176
xN8 Log % RDA Vitamin A 4.5 4.6 15.2 0.015
xN9 % RDA Thiamin 105.7 117.5 39.8 0.01?
xN10 % RDA Riboflavin 122.1 139.0 50.1 0.01’
xNH % RDA Vitamin 86 68.9 80.0 50.1 0.01*
lez Log % RDA Vitamin 812 2.3 2.4 20.3 0.0176
leB % RDA Vitamin C 139.4 157.6 18.1 0.011‘
xN14 % Recommended Energy 80.2 86.1 28.3 0.01*

Intake

 

* The N reflects the sample which had no missing values for each of the
indicators of demographic characteristics of the 10% random subsample
generated from the total study population (N=2406).

+ The units of measure for each indicator are listed in Appendix A-4.
# p §_.O1 .

99

TABLE 34. Group Means and Tests for Group Differences For Indicators of
Food Related Behaviors by the Dependent Variable, Problem
Group (N=2,385)

 

 

 

 

Problem Groups ¢g+
Food Related Behavior No
Indicators Problem Problem F Ratio p

x“ Weight Status 1.1 1.1 5.5 .05‘
xF2 # Mentions Milk 4.2 4.7 11.7 .01"6
xF3 # Mentions Coffee/Tea 3.9 3.0 30.5 .015
xF4 # Mentions Meat 3.1 3.4 15.7 .016
xF5 # Mentions Legumes 0.3 0.3 8.7 .016
xF6 # Mentions Starch 1.9 2.2 27.2 .016
xF7 # Mentions Vegetables 3.6 3.8 4.7 .05’
xF8 # Mentions Bread 4.3 4.7 17.5 .015
ng # Mentions Cereal 1.0 1.3 25.6 .015
XFlO # Mentions Frozen Desserts 0.4 0.5 7.0 .01‘5
xF1] # Mentions Cakes, Pies 1.7 1.8 4.0 .0576
XFlZ # Mentions Butter/Margarine 1.7 1.9 5.8 05*
xF13 # Mentions Nut Butter 0.2 0 3 12.4 .016
xF14 # Ate Alone-Day 2 0.1 0.1 13.4 .016
XF15 # Days Ate Breakfast 2.5 2.7 31.2 .01‘8
XF16 # Days Ate Lunch 2.3 2.5 11.7 .016
xF17 # Days Ate Supper 1.7 1.9 8.5 .01‘5

 

* The N reflects the sample which had no missing values for each of
the indicators of food related behaviors of the 10% random sub-
sample generated from the total study population (N=2406).

+ The units of measure for each indicator are listed in Appendix A-4.
f

6

p §_.05 .
p‘5_.01 .

100

indicators included 'hi the model. The column labeled maximum/minimum
variance is the ratio of maximum to minimum variance determined from
ANOVA's (analyses of within group variances). In the ANOVA's each of the
indicators was the dependent variable and the problem versus no problem
variable was the independent variable. Tables 32-34 list the group means
and results of tests for group differences (F ratio) for each of the three
sets of independent variables. Each of the independent variable sets will

be discussed separately combining the results from Tables 31-34.

Demographic Descriptors

Demographic characteristics have consistently been demonstrated to
significantly affect nutrient intakes or food behaviors (Lund and Burk,
1969; Eppright et al., 1970; Sims, 1972; Yetley, 1974; and Adrian and
Daniels, 1976). Age (X 01), sex (X 02), race (X 03), education levels of
male (X 04) and female (X 05) heads of households, poverty level (X 06)
and city characteristics of respondents place of residence (City C1 =
central city X 07 and City C2 = suburban X 08) were used as indicators of
demographic characteristics in the model.

The mean age of the sample was 31 years (see Table 31). The mean age
was 35 years for the problem group and 28 years for the no problem group
(p:f.01). This result was expected given the findings concluded in Phase
I. The none category had more children than any of the other intake
factor categories. There were more women in the problem category and more
men in the no problem category (p51.01). Both groups had more whites than
blacks, but this was reflected in the total NFCS sample population. In
education level of male and female heads of households the problem and no
problem groups were similar. In both cases the females had a total sample

average (R = 12.1 years of schooling) of approximately one year less of

101

education than the males (i = 13.0 years of schooling). Poverty level and

city characteristics were similar for the problem groups.

Nutrient Intake

The means for the total sample and each of the problem groups for the
percent of calories from the macronutrients fats (i = approximately 40%)
and carbohydrates (2 = approximately 44%) were similar. This was expected
given the results from Phase I for the PFC index and composite index (see
Chapter IV). The factor intake groups had similar proportions in the PFC
categories. By subtracting the mean total percent calories of fats and
carbohydrates from 100 percent, the mean total percent calories from
protein was calculated (i = approximately 16%). In 1976, Page and Friend
estimated the ratio of the macronutrients in the U.S. diets to be
approximately: protein (12%), fats (42%), and carbohydrates (46%). Our
mean intakes estimated from a random subsample of the NFCS were slightly
different: protein (14%), fats (40%), and carbohydrates (44%). Our
findings were comparable to the estimates Hegsted (1979) reported from the
NFCS. Hegsted (1979) reported percent of calories from fats to be
decreasing to 39-42 percent as estimated from the NFCS compared to the
1965 National Food Consumption Survey data estimate of 42-45 percent. Our
findings supported the trend Hegsted noted by saying "the message promoted
by the federal government to reduce fat intake has reached segments of the
U.S. population." The Legislative Branch has promoted the decreased
intake of fat in “Dietary Goals I and II" (U.S. Senate Select Committee on
Nutrition and Human Needs, 1977 a. and 1977 0.). The Executuve Branch
has been responsible for deve10ping and distributing several publications
suggesting the U.S. population decrease intake of fat. Examples of these

publications include: "Healthy People" (Surgeon General U.S. Department

102

of Health, Education and Welfare, 1979) and "Dietary Guidelines for
Americans" (USDA and Department of Health, Education and Welfare (1980).

Protein has repeatedly been reported as comprising 10-12 percent of
the calories ingested (Hegsted, 1979). Protein intake has been reported
to be increasing and our data supported this trend.

The total intake of calories has been reported as decreasing
(Hegsted, 1979). The mean total intake from the sample was estimated as
83 percent of the REI, Recommended Energy Intakes (NAS, Food and Nutrition
Board, 1980). The mean for the problem group (i = 80%) was lower than the
no problem group (i = 86%) (R:.10) (see Tables 31 and 33). Tables 31 and
33 further illustrate that the mean intakes for all indicator nutrients
were higher for the no problem category (p35 .01). This may be due to the
fact that they are just eating more than the problem category on the
average as illustrated by the REI.

The mean values obtained from three day average percentages of the
RDA for nutrient intakes were used to determine a rank order for average
percent intakes of nutrients. The rank order (largest to smallest)
derived was: protein (X = 167%), vitamin 812 (i = 158%), vitamin C (i =
149%), phosphorous (i = 135%), riboflavin (i = 131%), thiamin (i = 112%),
iron (2 = 102%), calcium (2 = 87%), magnesium (i = 81%), and vitamin B6 (2
= 74%). From these data, no nutrients on the average fell below the 59.9
percent of the RDA used in deriving the marginal nutrients. Averages or
means masked the ranges and numbers of individuals which fell in the low
end of the percent of RDA intakes for specific nutrients. From these
data, calcium, magnesium and vitamin 86 had the lowest mean values for the
percentages of RDA which was similar to the findings of Phase I (see
Chapter IV).

To use these averages for any kind of policy or nutrition education

103

programs would not be recommended. To determine accurately specific
nutrients which are marginal in terms of the RDA for the 0.5. population,
intakes would need to be determined for the same age/sex groups as are the

RDA.

Food Related Behaviors

Table 31 and 34 show the total group and problem group means for
indicators of food related behaviors. Weight status averaged at 10
percent above the NAS, Food and Nutrition Board (1980) recommended weight
ranges. See Chapter III for description of variable constructed to
represent weight status. Given the results of Phase I regarding weight
status (see Chapter IV) the findings in Phase II were expected. For all
food group total number of mentions for three days (expect coffee and
tea), the no problem category had larger average numbers. This was
expected since the no problem group had larger percentage intakes of the
REI (see Table 33). From these data based on mean (2) total number of
mentions for the three days, bread (i = 4.5), milk (i = 4.4), vegetables
(i = 3.7), and meat (i = 3.2) had the most mentions by respondents. In
investigating the number of days specific meals were eaten (XF15 - XF17)’
the no problem group had more meals eaten than the problem group. These
data supplement the findings from Phase I on specific meals eaten to
contribute to the derived meal patterns. Supper (R = 1.8) and lunch (3 =
2.4) were the meals most often skipped based on the average total number
of days mentioned. Breakfast (i = 2.6) was the meal which had the largest

mean for total number of days eaten in both groups.

104

Tests for Multicollinearity

For purposes of assessment and clarification in interpretation of
results, tests for multicollinearity were applied for indicators of the
independent variable sets. Linear independence is assumed among
variables in general linear model (GLM) analysis techniques. It should be
stated that estimates may be derived and statistical tests performed on
data sets where Inulticollinearity exists. However, the 'results of
estimates and statistical tests where perfect linearity exists are not
readily interpretable or stable.

To determine the extent of multicollinearity (for description see
Chapter 111) within each of the independent variable sets the determinant
of the correlation matrices was derived and principal component factor
analyses were conducted (see Appendix 8, Tables BI-B6). SPSS Subprogram
Factor (Nie et al., 1975) was used to conduct tests for multicollinearity.
The 10 percent random subsample (N = 2,406) originally selected from the
study population was randomly split into two 50 percent subsamples
(N = 1,203) to cross-validate results obtained.

In Tables BI-B3 (see Appendix B) the: correlation Inatrices and
calculated determinants (det.) for each of the independent variable set
indicators are shown: demographic characteristics (det. = .4); nutrient
quality (det. = .1); and food related behaviors (det. = .2). A
determinant larger than zero indicated some linear independence among
variables. The nutrient quality indicators had the highest level of
multicollinearity (closest to zero) which was expected given the high
correlations (r = 3:.7) between many of the variables and the fact that
many foods contain similar sets of nutrients. Pennington (1976) and
Jenkins (1982) investigated the correlations between nutrients. Beyond

the criteria specified in Chapter III non-perfect multicollinearity ( a

105

non-zero determinant) was the only criterion used for incorporating a
particular set of indicators in our independent variable sets. Each of
the three sets met this criterion.

None of the indicators. of demographic characteristics and ‘food
related behaviors were highly correlated. Moderate correlations existed
between the demographic characteristics: race and central city (r = .4);
female head of households educational level and poverty level (r = .4);
and central city and suburban (r = -.5). The only indicators of food
related behaviors that were moderately correlated (r = .4 -.6) were the
meat and bread food groups (r = .4). As was indicated from the
determinant of the correlation matrix several of the indicators of
nutrient quality were highly correlated (r = 53.7). The indicators which
were highly correlated are listed in Table 35.

The second set of analyses to determine linear independency among
indicators of the independent variable sets were principal component
factor analyses. The principal component solution was derived from
standardizing the variance-covariance matrix resulting in a correlation
matrix (Tables B4-B6, Appendix B). This correlation matrix was a "matrix
of cosines where each entry eij represented the cosine of the angle
between the variable vectors 1 and 521 (Woelfel et al., 1980). This
correlation matrix was then orthogonally decomposed and rotated to yield
eigenvectors and eigenvalues. For each analysis, a set of reference
coordinates or axes were placed upon the sets of variables and projections
were measured (loadings =x ) for each variable on the axes.

Tables 84-86, Appendix 8 illustrate the results of the principal
component analyses in terms of the rotated (transformed) factor
structures presented. These tables showed that each of the major factors

or axes determined in the factor eigenanalyses of the three sets of

106

TABLE 35.

Identification of Highly Correlated (RC:,7) Indicators of Nutrient
Quality (N=1177)*

 

 

 

Indicators of Nutrient Quality __jg;____
% Calories from Fats - % Calories from Carbohydrates .81
% RDA Protein - % RDA Calcium .68
% RDA Protein - % RDA Magnesium .70
% RDA Protein - % RDA Vitamin A .75
% RDA Protein - % RDA Riboflavin .72
% RDA Protein - % RDA Vitamin 85 .76
% RDA Protein - % Recommended Energy Intake .73
% RDA Calcium - % RDA Magnesium .70
% RDA Calcium - % RDA Phosphorous .85
% RDA Calcium - % RDA Riboflavin .79
% RDA Iron - % RDA Phosphorous .72
% RDA Magnesium - % RDA Phosphorous .77
% RDA Magnesium - % RDA Riboflavin .71
% RDA Magnesium - % RDA Vitamin B5 1 .69
% RDA Magnesium - % Recommended Energy Intake .71
% RDA Phosphorous - % RDA Riboflavin .68
% RDA Phosphorous - % Recommended Energy Intake .71
% RDA Thiamin - % RDA Riboflavin .68
% RDA Thiamin - % RDA Vitamin B6 .69
% RDA Thiamin - % Recommended Energy Intake .70
% RDA Riboflavin - % RDA Vitamin B6 .74
% RDA Riboflavin - % RDA Vitamin B12 .71
% RDA Riboflavin - % Recommended Energy Intake .68

 

* Sample size reflects results of case elimination due to missing
data from a 5% subsample of total study population (N=24,362,
5%=1203). Two 5% subsamples were determined and cross validated.

107

independent variables were primarily explained by one of the indicators in
the set. For example, in the principal component analyses of the
demographic characteristics Factor 1, race ( A = .99) explained the
largest proportion of the variance. For Factor 2, suburban ( A = .96)
explained the largest proportion of variance and in Factor 3 it was
educational level of the male head of household ()1 = .97). This pattern
was consistent across each of the three sets of indicators. Factor
analyses were used to identify relative linear independency among
indicator variables. In this sense, factor analyses were used for
exploratory purposes (exploring and detecting patterns in variables) (Nie
et al., 1975:17) and not for purposes of estimation or statistical
testing.

Based on the findings from the determinants of the correlation
matrices and the principal component analyses, the researchers. were
satisfied that the indicators as represented in Figure 4 (see descriptions
in Appendix A-4) were relatively linear independent.

Once the indicators for the three sets of independent variables had
been determined, correlation matrices were derived for each of the three
independent variable sets with one another (see Appendix B, Tables B7-B9).
Most correlation coefficients were low, however, a moderate, negative
relationship existed between sex and iron (r = -.5) in the matrix of
demographic characteristics and nutrient quality indicators. Given the
higher RDA standard for women and the lower percentage of women meeting
the standard, this was an expected result. In the matrix determined
between nutrient quality and food related behaviors, several moderate
correlations were determined. The milk food group was moderately,
positively correlated with percent RDA calcium and percent RDA riboflavin

(r = .6, each), percent RDA magnesium (r = .5), percent RDA phosphorous,

108

FIGURE 4. Model Estimated To Assess Interrelationships
Between Sets of Variables Affecting Food
Related Behaviors. Appendix A-4 Defines
Variables Included in the Model.

109

.e $5un

      
          

v
a on. our a . ... a.» 3.
cm .1:

...... r...
in! 3‘
...... 30.22%. oz:
can amid...

a. OOOE {a
5‘ an
t. a

we «I

noun .8 a

..a $5.2. . 2:3 ..
x: 5.89:. .6
ex zozaofifizme.

O.:a<c@02wo

110

percent RDA vitamin B6 and percent RDA vitamin 812 (r = .4, each). The
RDA percent for vitamin A was moderately, positively correlated to the
total number of mentions of the vegetable food group (r = .4). The food
group representing total number of mentions of cereal was moderately,
positively correlated to percent RDA riboflavin and percent RDA vitamin B6

(r = .4, each).

Discriminant Analyses

 

With the indicators determined for the three sets of independent
variables, discriminant analyses were computed to: (1) produce'a check on
homogeneity in the covariance matrices for each of the independent
variable sets and (2) determine the relationship between each of the three
sets and the dependent variable, identification of problem with intake.
These results are presented separately.

As a final quantitative check on multicollinearity and to assess an
underlying assumption of discriminant analysis (see Chapter III), the log
determinants of the covariance matrices were derived for each of the
discriminant analyses run (see Table 36). The rank and log determinants
were calculated for the problem versus no problem dependent variable
groups within each of the three sets of independent variables. Each of
the covariance matrices were full in rank (see Chapter III).

The rank of a matrix is the order of the largest square submatrix
whose determinant does not equal zero. To meet full rank no column vector
can be linearly dependent upon another column (the same holds true for
rows) and the number of columns must exceed the number of rows. When
linear dependency exists (as when multicollinearity exists) the
dimensionality of space needed to represent the variable vectors is

deficient (geometric definition of rank). In this situation one or more

111

of the variable vectors is highly correlated and in the case of perfect
multicollinearity the vectors are colinear.

The log determinants were obtained to check for two criterion
regarding the covariance matrices: (1) non-singular results and (2)
similar determinants for each of the problem groups within each of the
independent sets. The SPSS Subprogram Discriminant (Nie et al., 1975)
actually prints the message "singular further analyses can not be
conducted" when singular results are obtained. Given the two criterion
(full rank and non-singular determinants), the data were determined to
satisfy the assumptions of discriminant analyses. Subsequently,
discriminant analyses were performed to determine canonical discriminant
functions for the dependent variable and each of the independent variable
sets.

Discriminant analyses were used to determine the relationship
between each of the independent variable sets and the dependent variable.
The output from these analyses was a canonical correlation coefficient
(RC) which represented the correlation between the dependent variable and
the s_et_ of independent variables being assessed. The square of the
canonical correlation (or the eigenvalue, RC2) represents the amount of
variance shared by the two sets. The results of these analyses are
presented in Table 37. The canonical correlations derived for each of the
groups were quite similar: demographic (Rc = .25), nutrient quality (Rc =
.23), and food related behaviors (Rc = .27). These were low'correlations.
These canonical correlations summarized the strength of association
between the variable sets and also provided a means for comparing the
association across groups. The eigenvalues (Rcz's) were a measure of the

proportion of variance shared by variable sets. The Rcz's were small,

demographic and nutrient quality (RC2 = .06) and food related behaviors

112

TABLE 36.

Tests For the Equality of Group Covariance Matrices For Each of the
3 Independent Variable Sets and the Dependent Variable (N=2406)*

 

 

 

Independent Dependent
Variable Problem Log
Set Group Rank Determinant

Demographic Yes 8 4.5

No 8 4.6
Nutrient Yes 14 70.5
Quality No 14 70.6
Food Related Yes 17 4.4
Behaviors No 17 3.3

 

* The sample size included in each of the analyses varied as a result
of case elimination due to missing data. These analyses were run
on two 5% subsamples of the 10% (N=2406) and the results were cross
validated.

TABLE 37.

Canonical Discriminant Functions For the Dependent Variable By the
3 Sets of Independent Variables (N=2406)*

 

 

 

Canonical Chi

Independent Eigenvggue Correlation Squgre

' *
Variable Set N (Rc ) (RC) (x ) df p
Demographic 1809 .06 .25 59 8 '000
Nutrient
Quality 2356 .06 .23 63 14 .000
Food Related 2385 .08 .27 89 17 .000

Behaviors

 

* Sample size varies as a result of case elimination due to missing data.
These analyses were run on two separate 5% subsamples of the 10%
random subsample generated from the total study population (24,362).

113

(RC2 == .08). In each case over 90 percent of the variation in the
disciminant space remained unaccounted by the between group differences.
These results may mean several things: (1) the problem versus no problem
groups were very similar in composition or patterned variation and
differentiation or explanation may have been difficult; (2) the
indicators used to represent each of the independent variable groups were
not appropriate to differentiate accurately the problem groups; (3) non-

linear relationships and/or interaction affects existed between

variables; and/or (4) gross unreliability existed in the data set.

Canonical Analyses

The final task in Phase II determined the quantitative relationships
between the three sets of independent variables by deriving canonical
correlations (Rc's) and canonical variate loadings ( A '5) (see Chapter
III). From the eigenanalyses, Tables 38-40 were produced which list each
of the canonical variates derived with the calculated canonical
correlations (Rc's) and eigenvalues (Rcz's). Figure 5 represents the
model estimated based on the jjggt canonical variates derived for each
group and the discriminant canonical analyses from the previous section in
this chapter. Tables 41-43 list the canonical variate loadings ( A '5)
showing which variables accounted for the most variation in the underlying
canonical variates. For example, the first reference axis derived,
canonical variate l, for each set accounted for the greatest proportion of
variance between the two sets. Between each of the sets the Rcz's
(eigenvalues) were moderate: demographic-nutrient quality (RC2 = .5),
demographic-food related behaviors (RC2 = .5) and nutrient quality-food

2

behaviors (Rc = .6). These can be interpreted to mean that 50-60 percent

of the variation between the two groups was explained by the specified

114

TABLE 38.

Canonical Correlations Among the Indicators of Demographic Characteristics
With the Indicators of Nutrient Quality (N=2406)*

 

 

 

 

Canonical Nilk's Chi

Canonical Eigenvalue Correlation Lambda Square

Variate (RQZ) (BE) (A) (X?) df p
1 .54 .74 .21 1375 112 .00+
2 .37 .61 .46 684 91 .00+
3 .13 , .36 .73 274 72 .00+
4 .08 .29 .84 150 55 .00+
5 .04 .19 .92 75 40 .01+
6 .02 .16 .95 42 27 .07
7 .02 .12 .98 20 16 .22

8 .01 .08 .99 6 7 .51

 

115

TABLE 39.

Canonical Correlations Among the Indicators of Demographic Characteristics
With the Indicators of Food Behaviors (N=2406)*

 

 

 

 

Canonical Nilk's Chi

Canonical Eigenvalue Correlation Lambda Square

Variate (Re?) (Rc) (A') (X?) df p
1 .47 .68 .29 1096 136 .00+
2 .22 .47 .55 534 112 .00+
3 .14 .38 .70 313 90 .00+
4 .08 .28 .82 175 70 .00+
5 .06 .25 .89 101 52 .00+
6 .03 .17 .95 46 36 .13
7 .01 .12 .98 20 22 .61
8 .01 .09 .99 7 10 .74

 

* The sample size varies as a result of case elimination due to missing
data. These analyses were run on two separate 5% subsamples of the 10%
random subsample generated from the study population (N=24,362).

+ P 5_.0l.

116

TABLE 40.

Canonical Correlations Among the Indicators of Nutrient Quality With
The Indicators of Food Behaviors (N=2406)*

 

 

 

Canonical Wilk's Chi

Canonical Eigenvalue Correlation Lambda Square

Variate (RE) (Re) (A ) ()3) df p
1 .64 .80 .04 3776 238 .00+
2 .43 .65 .11 2593 208 .00+
3 .36 .60 .19 1947 180 .00+
4 .30 .55 .29 1428 154 .00+
5 .27 .52 .42 1013 130 .00+
6 .20 .45 .57 653 108 .00+
7 .12 .35 .72 388 88 .00+
8 .09 .29 .81 241 70 .00+
9 .05 .22 .89 135 54 .00+
10 .03 .18 .94 76 40 .00+
11 .02 .13 .97 37 28 .05’
12 .01 .11 .98 18 18 .48
13 .00 .06 1.00 5 10 .91
14 .00 .03 1.00 1 4 .91

 

* The sample size varies as a result of case elimination due to missing
data. These analyses were run on two separate 5% subsamples of the 10%
random subsample generated from the study population (N=24,362).

+ p'g .Ol.
f p §_.05.

FIGURE 5.

117

Estimated Model Based on First Canonical Variate Analyses.
Estimates Presented are Canonical Correlations Between the
Variable Sets. Tables 41 — 43 List the Canonical Variate

Loadings for Indicators of Independent Variable Sets.

 

118

.m mma¢mm

      
          

v o
0 ID. x N a m. H N— r g} i
3; re
(a :e
an. ...
.3. «3.22.3 ...x
a: 3:32.
... 902 ....f
:5. fl
..bl .I
..a mafia-2:3 .
x e 5.50.... .o
..x 20.20.6323.
0
A. u 6 N. u m

OEI‘KGOSNO

119

TABLE 41.

Canonical Variate Loadings For the First 5 Canonical Variates Determined
Between)the Indicators of Demographic Characteristics and Nutrient Quality
(N=l782 *

 

 

Canonical Variate Loadings (A)

 

 

Indicator l 2 3 4 5

XDl Age .89 .43 .06 -.ll .05
XDZ Sex -.36 .9l -.l4 .01 -.ll
XD3 Race -.l4 .05 .78 .36 .03
XD4 Education Male -.ll .l9 -.29 .39 .77
XDS Education Female .12 -.l4 -.7l .42 .09
XD6 Poverty Level . .l8 .00 -.4l .68 -.39

Living Locale:

XD7 Central City .03 .03 .31 .53 .22
X08 Suburban -.0l -.l3 -.32 -.2l -.2l
XNl % Calories Fats .l3 -.l3 -.06 -.04 .l0
XN2 % Calories CHO -.25 .06 -.ll -.2l .3l
XN3 RDA Protein -.32 -.52 .08 .23 .02
XN4 RDA Calcium -.08 -.52 -.43 .01 .ll
XN5 RDA Iron .5l -.57 .l7 -.09 .l3
XN6 RDA Magnesium -.04 -.33 -.29 -.03 .02
XN7 RDA Phosphorous .23 -.54 -.27 .l3 -.03
XN8 RDA Vitamin A -.02 -.04 -.05 .32 .30
XN9 RDA Thiamin -.l2 -.24 .24 .06 .29
leD RDA Riboflavin -.26 -.32 -.l5 .03 .05
lel RDA Vitamin B6 -.18 -.58 .13 .23 .06
leZ RDA Vitamin 312 -.l2 -.59 -.l7 .32 -.Zl
XN13 RDA Vitamin C -.08 -.l7 -.l3 .53 .52
XNl4 Recommended Energy .04 -.20 -.ll .l9 .22

Intakes

 

* The sample size varied due to the missing values for each of the sets.
These analyses were conducted on two 5% subsamples of the l0% random
subsample.

120

TABLE 42.

Canonical Variate Loadings For the First 5 Canonical Variates Determined
Between the Indicators of Demographic Characteristics and Food Related
Behaviors (N=1807)*

 

 

Canonical Variate Loadings (A);

 

 

Indicator l 2 3 4 5

X01 Age .97 -.17 -.03 .07 -.02
XDZ Sex .l9 .24 .60 -.3l -.6l
XD3 Race -.l7 -.45 .53 .59 .02
XD4 Education Male .09 .54 .44 -.ll .64
XDS Education Female -.09 .65 -.24 -.lO .l5
X06 Poverty Level .l6 .48 -.4l .38 -.l2

Living Locale:

XD7 Cl - Central City -.02 -.Ol .50 .29 .04
X08 C2 - Suburban -.07 .53 -.22 .33 -.14
XF] Weight Status .22 -.l6 .l0 .04 -.37
XF2 Milk -.30 .l8 -.l7 -.23 .35
XF3 Coffee/Tea .80 .04 -.38 -.03 -.l3
XF4 Meat .02 -.07 -.4l .5l .23
XF5 Legume -.09 -.37 -.08 -.04 .02
XF6 Starchy Dishes -.ll -.44 -.06 .47 .29
XF7 Vegetables .39 .l2 -.l7 .l2 .27
XF8 Bread .05 -.32 -.44 .10 .29
XF9 Cereal -.Ol -.03 -.l3 -.22 .48
XF10 Frozen Desserts -.l8 .41 -.20 -.23 .28
XF11 Cakes, Cookies -.l6 .39 -.35 -.23 .41
XFlZ Butter & Margarine .16 .03 -.34 .l4 .Ol
XF13 Nut Butters .29 .07 -.l6 - 07 .24
XF14 Ate Alone .43 .02 .67 -.T7 .43
XF15 # Days Breakfast .10 -.09 -.l3 .06 .26
xFl6 # Days Lunch .lO -.50 -.22 -.33 .ll
XF17 # Days Supper .03 -.50 -.lZ -.56 .Ol

 

* The sample size varied from 2406 as a result of case elimination due to
missing data. These analyses were run on two separate 5% subsamples of the
10% random subsample generated from the total study population (N=24,362).

TABLE 43.

Canonical Variate Loadings for the First 5 Canonical Variates Determined

121

Between the Indicators of Nutrient Quality and Food Related Behaviors

 

 

 

(N=2357)*
Canonical Variate Loadings (X)

Indicator l 2 3 4 5
X N1 % Calories Fat -.22 .43 -.34 -.18 .30
X N2 % Calories CHO .35 -.17 .37 .16 -.37
X N3 RDA Protein .49 .55 -.10 -.12 .05
X N4 RDA Calcium .73 .35 .06 .06 .19
X N5 RDA Iron .10 .60 .32 -.13 -.06
X N6 RDA Magnesium .45 .52 .54 .05 .12
X N7 RDA Phosphorous .37 .55 .12 .01 .14
X N8 RDA Vitamin A .36 .34 .34 -.28 .25
X N9 RDA Thiamin .42 .58 .17 -.24 -.10
X N10 Riboflavin .76 .40 .ll -.12 .03
X N11 RDA Vitamin B 6 .55 .37 .19 -.31 -.08
X N12 RDA Vitamin B 12 .50 .41 -.16 -.15 .12
X N13 RDA Vitamin C .29 .23 .15 -.26 .05
X N14 Recommended Energy

Intakes .30 .82 .05 .00 .02
X Fl Weight Status -.10 .00 .02 .01 .03
X F2 Milk .83 .12 .14 .08 .18
X F3 Coffee/Tea .56 .22 .54 .17 .20
X F4 Meat .12 .69 -.04 -.23 .05
X F5 Legume .00 .07 .26 .06 -.14
X F6 Starchy Dishes .05 .36 -.08 -.10 -.15
X F7 Vegetables -.02 .37 .29 -.27 .33
X F8 Bread .10 .62 -.03 .10 -.01
X F9 Cereal .53 -.05 .52 -.24 -.17
X F10 Frozen Desserts .29 .24 -.04 .02 .08
X Fll Cakes, Cookies .28 .47 .12 .16 -.09
X F12 Butter & Margarine -.02 .51 .08 .00 .19
X F13 Nut Butters .25 .06 .24 .25 .01
X F14 Ate Alone -.14 -.09 .05 .01 .01
X F15 # Days Ate Breakfast .24 .25 .19 -.10 .01
X F16 # Days Ate Lunch -.05 .08 .02 -.02 -.04
X F17 # Days Ate Supper .07 .08 .07 .09 -.13

 

* Sample size varied for the 2406 as a result of case elimination due to
missing data. These analyses were run in two separate 5% subsamples
of the 10% random subsample generated from the total study population
(N=24,362).

122

indicators. In the demographic-nutrient quality analysis, age ( A = .97)
of the demographic indicators and iron ( A = .51) in the nutrient quality
indicators were the specific. variables primarily' accounting 'for ‘the
variance explained by the first reference axis (see Table 41). In the
demographic-food related behaviors, age ( A = .97) as the demographic
indicator and coffee/tea ( A.= .80) as the food related behavior indicator
were the indicators which primarily accounted for the proportion of
variance explained by the two sets of indicators (see Table 42). The
average three day percent RDA for riboflavin ( A = .76) and calcium
( A = .73) of the nutrient quality indicators and the total number of
mentions of the milk group ( A = .83) in the food related behavior
indicators were the variables which primarily accounted for the variance
explained by the first canonical variate between the groups (see Table
43).

From the Tables 41-43, the specific indicators which accounted for
the variance explained by the canonical variates determined between
groups can be determined. Many of these results were expected. For
instance, in the analyses between the nutrient quality and food related
behaviors group, percent RDA riboflavin and calcium and milk were the
indicators with the highest loadings. Milk is a good source and provides
large portions of the RDA requirements for riboflavin and calcium. In the
second orthogonal canonical variate derived between those groups, REI and
bread and meat were the primary indicators accounting for the explained
variance between the groups. Bread and meat had among the highest average
number of mentions for the three days and were two food groups eaten by
the most respondents, which would explain the relationship. In the
analyses between demographic characteristics and nutrient quality and

food related behaviors, age was the indicator which accounted for the

123

primary variance explained by the first canonical variates in both cases.
This was expected because eating behaviors and RDA vary significantly

between age categories (Crocetti and Guthrie, 1982).

Summar

Phase II of this investigation was designed to investigate and
quantify the relationships between sets of independent variables
representing demographic characteristics, nutrient intake, and food
related behaviors and the dependent variable, problem versus no problem
group. Multivariate analyses were used to investigate factors
interacting in food related behaviors.

The no problem group was younger, ate more in terms of total number
of mentions of specific food groups, and ate more (in number) of specific
meals than the group that identified some type of factor affecting or
problem with their intake. Many of the other findings as the no problem
group having all average percent RDA intakes higher than the problem group
were expected given the larger intakes in terms of calories and specific
food groups of the no problem group over the problem group.

Discriminant analyses were used to quantify the relationship between
each of the sets of independent variables and the dependent variable. Low
correlation coefficients were determined (chf.3). These results may
have meant that the problem no problem groups were very similar in
composition or patterned variation and subsequently differentiation or
explanation may have been difficult. It may also have meant that the
indicators used in the model to represent each of the sets of independent
variables were not appropriate, not measured well, and/or misspecified in
functional form .

Canonical analyses were conducted to determine the relationships

124

between each of the sets of indicators representing independent
variables. The correlations obtained from canonical variate analyses
(see Figure 5 and Tables 38-40) were moderately high (Rc_3.7-.8) and the
amount of patterned variation accounted for by the sets of variables
moderate (50 to 60% in the first canonical variates). Age was the
indicator explaining the largest proportion of the variation between sets
when demographic characteristics were included. Relationships between
specific nutrient average percent RDA intake and food groups were
documented (such as calcium and riboflavin and the milk group).

This investigation further "unraveled“ specific relationships
between sets of variables affecting food related behaviors. It built upon
previous findings which identified relationships between these
independent variable sets. It went a step beyond traditional econometric
techniques, in that the correlations were derived between sets rather than
individual indicators. The findings may be used to further refine and
direct future analyses to unfold a better set of measured indicators for
each of the independent variable sets. Analyses to derive eXplanatory
variables are critical steps in groundng understandings of conceptual
schemes which facilitate eventual theory development in applied

nutrition.

125

FOOTNOTES
1As Woelfel et al., 1980 point out, the diagonal entries of the
standardized correlation matirx are l or unity. This is due to the

fact that where i = 3 the angle is 0 and the cosine of 0 = 1.

CHAPTER VI
SUMMARY AND CONCLUSIONS

The Department of Agriculture (USDA) staff included two questions in
the 1977-78 Nationwide Food Consumption Survey (NFCS) on the U.S.
population's behavior that could affect food and nutrient intake
patterns. First, individuals participating in the NFCS reported if they
were on a doctor prescribed diet, on a group, or an individual diet
regimen. They then responded to a list of nine items that might affect
food consumption. These nine items were: being on a diet to lose weight;
being on a diet to gain weight; having a chewing problem; having a medical
problem like diabetes or an allergy; foods not agreeing with them; not
liking to eat breakfast early in the morning; having no interest in
cooking for one person; not liking to eat certain foods; and other.

Almost 50 percent (49.5%) of the NFCS respondents were on a diet
listed in the first question or checked at least one item in the second
question (Crocetti and Guthrie, 1982). Thus a large sample identified
themselves as having at least one behavior that potentially could affect
dietary intake.

This secondary analysis of NFCS data was conducted in two phases. In
Phase I the entire study population (N = 24,362) was divided into four
groups based upon whether answers to the two questions were medically or
Inonmedically related: MEDICAL (4.0%), NON-MEDICAL (39.5%), NON-MEDICAL &
IMEDICAL (6.0%), and NONE (50.5%). The MEDICAL group reported either a

126

 

127

doctor prescribed diet and/or a medical problem like diabetes or an
allergy. The NON-MEDICAL group reported being on a group or individual
diet regimen and/or any one of the other eight factors affecting intake
(excluding the one medical problem). The NON-MEDICAL & MEDICAL group
reported a medical problem(s) and a non-medical factor(s). The NONE group
did not report any of the medical problems or non-medical factors. Sets
of descriptive characteristics were derived for each of the four groups.
The statistically determined descriptions included: demographic
characteristics; meal and snack patterns; nutrient quality assessment;
and food related behaviors.

The NON-MEDICAL & MEDICAL and MEDICAL groups had the most respondents
over 55 years of age and were comparable in socio-economic
characteristics. As expected, they lived in smaller households with more:
non-working adults, female only headed households, and lower education
levels for head of households. These groups also were similar in
frequency of dietary intake from food groups and in eating out and eating
alone behaviors.

The NONE and NON-MEDICAL had the most respondents under 18 years of
age, living in larger households, and with employed head(s) of household.
Personal food behaviors such as intake frequency for food groups, eating
alone, and eating out were alike for these two groups.

The NONE and MEDICAL groups had the most respondents eating three-
meal-a-day patterns. The NON-MEDICAL and NON-MEDICAL & MEDICAL had the
most respondents reporting patterns of two and/or three meals a day. All
groups reported snacking ( 373%) with groups not varying considerably
from one another.

The NON-MEDICAL and NON-MEDICAL & MEDICAL had more respondents

ingesting foods meeting less than 60 percent of their Recommended Dietary

128

Allowances (RDA) for seven nutrients: calcium, iron, magnesium, vitamin
A, vitamin B6’ vitamin 812, and vitamin C (marginality index).
Conversely, the NONE and MEDICAL groups had the most respondents ingesting
foods with more than 59.9 percent of the RDA for the seven nutrients.

Therefore, meal patterns and total number of mentions of food groups
were the characteristics which differentiated nutrient quality assessment
between the four groups. 1A descriptive composite for each groups is
presented. The application of these descriptive composites is discussed.

Phase II was a theoretical treatment of the data set to: 1)
demonstrate that if adequate data sets are available they could be used to
study factors affecting food behavior and 2) assess factors potentially
affecting food intake. The sample was divided into two groups, PROBLEM
versus NO PROBLEM affecting dietary intake. The respondents reporting any
one of the items in the two NFCS questions were in the PROBLEM group
(49.5%). Almodel was derived and estimated using multivariate techniques.
The dependent variable was the PROBLEM versus NO PROBLEM group. The three
independent variable sets represented demographic characteristics, food
related behaviors, and nutrient intakes. Correlational relationships
were determined between each of the variable sets.

Two five percent random subsamples (N = 1,200) of the total study
population used in Phase I were used to estimate the model and cross-
validate results. The preliminary tasks of Phase II consisted of
determining specific variables, indicator variables, from the NFCS data
set to define the independent variable sets in the model. For example,
the indicator variables determined to represent demographic
characteristics were: age, sex, race, education of male and female
head(s) of household, poverty level, and central city' and suburban

location of household residence. Also in the preliminary tasks was the

129

investigation of the underlying assumptions of the multivariate
statistical techniques used in analyses.

Descriptive statistics were generated for the PROBLEM versus NO
PROBLEM groups. The NO PROBLEM group was younger, ate more , and reported
more meals than the PROBLEM group. The NO PROBLEM group had higher mean
percent RDA nutrient intakes and mean caloric intakes than the PROBLEM
group.

These findings agreed with the findings of Phase I. On the average,
the respondents in the PROBLEM group reported fewer mentions of food
groups than the respondents in the NO PROBLEM group. They were also not
eating as many meals and had lower mean percentages for nutrient intakes
based on their RDA. Snacks did not contribute food groups or nutrients
needed for the PROBLEM group to equal food group or nutrient intake of the
NO PROBLEM group.

Once the indicator variables to be used in the model were described,
the model was estimated. [fiscriminant analyses were used to compute
correlation coefficients between the dependent variable, PROBLEM versus
NO PROBLEM, and each of the independent variable sets: demographic
characteristics, food related behaviors, and nutrient intake.
Correlation coefficients summarized the strength of association between
the variable sets and also provided a means for comparing association
across groups. Low correlations (RC1: .3) were determined between the
groups. The use of other indicators or more precise measurements of the
independent variable sets might explain more of the patterned variation in
the PROBLEM versus NO PROBLEM groups.

In addition, the» relationships between the sets of independent
variables were determined using canonical correlation analyses to

determine general strength of association between the sets and which

130

indicator variable(s) in the variable sets accounted for the most
patterned variation between sets. The correlations obtained for the first
canonical variate analyses were moderately high (Rc's of .7 to .8). The
RC2 represented the amount of patterned variation accounted for by the
sets of variables. The Rcz's were moderate (approximately 50 to 60%) by
the first canonical variates.

Age, sex, race, and poverty level were the characteristics
accounting for the largest proportions of patterned variation in the
analyses between demographic and nutrient quality indicators. Between
demographic and food related indicators age accounted for the most
patterned variation in the first canonical variate. A combination of
variables accounted for the patterned variation between nutrient quality
and food related indicators. In the first canonical variate, number of
mentions of the food group milk, percent RDA intake of riboflavin, and
percent RDA intake of calcium explained the most variation. This makes
sense since milk is a good source of riboflavin and calcium.

Findings from Phase 11 may be used to identify appropriate variables
to use as indicators for each of the independent variable sets. Also
shown was that with precise. measurements of appropriate variables,

multivariate analyses can be used to study factors affecting food intake.

Limitations of the Study

Limitations of conducting a secondary analysis of the NFCS are
discussed in detail in Chapter II. Considerations included: sampling
techniques, dietary data collection methodology, question selection and
wording, sample completion rates and variable construction. ‘These
considerations and financial constraints were viewed as the primary

limitations of this study.

131

Strengths of the study

 

The NFCS had approximately 25,000 respondents and data were
collected on demographics, food intake, and selected personal factors or
behaviors which may affect food habits. Descriptive composites were
derived based on respondents reported factors affecting dietary intakes.

The strengths of Phase II were the appropriate use of methodologies
to determine the relationships between sets of variables affecting or
interrelated in food behaviors and PROBLEM versus NO PROBLEM groups.
Estimation techniques used derived correlational/mathematical
relationships between sets of variables. Phase II was designed to have

theory building application.

Conclusions/Implications

Almost 50 percent of the NFCS sample responded "yes" to one or more
behaviors they had which could affect their dietary intake» Meal patterns
and food group intakes were the distinguishing characteristics in
nutrient quality assessment analyses among the four subgroups.
Mathematical relationships between sets of variables representing food

related behaviors were estimated in Phase II.

Phase I

Descriptive characteristics were determined for four' population
segments based upon their response to two dietary problem questions. The
medical problem groups have been and continue to be evaluated in other
national sample surveys, for example by the National Center for Health
.Statistics (1978). The population segments which reported non-medical
factors had not been evaluated in these specific terms in a national

isample prior to the NFCS. Expanded analyses and refinement of

132

measurements might provide researchers clearer understandings about
general public perceptions of factors affecting (positively or
negatively) their dietary intake. Communication specialists and
nutrition educators concur that messages presented in terms familiar and
meaningful to target populations should have greatest impact.

Phase I results complemented the findings of Crocetti and Guthrie
(1982). Meal patterns and total number of mentions of food groups were
the characteristics which differentiated nutrient quality assessment.
The two groups which had the largest percentages of respondents eating
three meals a day (the NONE and MEDICAL groups) had the smallest
percentages '01 the marginal (1:59°9% RDA) intake categories for seven
nutrients. The two groups identifying some type of non-medical factors
(NON-MEDICAL and NON-MEDICAL & MEDICAL) had the most respondents in the
nonregular three-meal-a-day patterns. These non-medical groups also had
the most respondents in the marginal intake categories for seven
nutrients.

In summary, the descriptive characteristics derived in Phase I can be
used to target nutrition education messages or strategies. For example,
the findings related to meal and snack patterns may alert nutritionists
and health educators to stress more nutrient dense food items since total
average caloric intake appears to be decreasing as determined by other
analyses of NFCS data.

The findings may also be used to define questions for future surveys
which seek to assess problems or factors affecting dietary intake. An
example of this would be redefining data collection on factors affecting
intakes and/or snacking behavior to improve possibilities for detailed

investigation.

133

Phase II

In Phase II multivariate techniques were explored in the estimation
of a model representing food related behaviors. Phase II results defined
the mathematical relationships between sets of variables interrelated in
food intake patterns.

The findings lead us to believe that multivariate analysis
techniques may be used to describe statistical associations between
variables affecting food intake. Several demographic variables as age,
sex, and race explained patterned variation between indicators of
demographics and nutrient intake and food related behaviors. The
significance of these specific variables has been documented in previous
investigations of food consumption data. However, the relationship of
these variables to the total model had not been investigated before.
Predictive understandings of food related behaviors may be furthered by
more precise measurement of variables and refinement in the process of
developing models. Use of analytical methods and theories from such
disciplines as sociology, anthropology, and psychology hold promise for

advancement in modeling of food related behaviors.

APPENDIX A
VARIABLE CONSTRUCTION AND DESCRIPTIONS

134

APPENDIX A-l

CALCULATIONS USED IN DEVELOPING THE "MARGINALITY SCORE" FOR SEVEN NUTRIENTS

 

 

 

 

 

 

"Weight" Percent RDA for

Seven Nutrients Assigned 3 Day Average

Calcium 1 3_80.0 (desirable)

Iron 2 3_60.0-79.9 (acceptable)

Magnesium 9 §_59.9 (marginal)

Vitamin A

Vitamin B6

Vitamin 812

Vitamin C

Raw # Nutrients Raw # Nutrients

Patterns Score Marginal Patterns Score Marginal
l l l l l l l 7 None 1 1 1 l 9 9 9 31 Three
1 l l l l l 2 8 l l l 2 9 9 9 32
l l l l l 2 2 9 l 1 2 2 9 9 9 33
l l l l 2 2 2 10 l 2 2 2 9 9 9 34
l l l 2 2 2 2 ll 2 2 2 2 9 9 9 35
l l 2 2 2 2 2 12
l 2 2 2 2 2 2 13 1 l l 9 9 9 9 39 Four
2 2 2 2 2 2 2 l4 1 l 2 9 9 9 9 40

l 2 2 9 9 9 9 41
l l l l l l 9 15 One 2 2 2 9 9 9 9 42
1 l l l l 2 9 l6
1 l l l 2 2 9 17 l l 9 9 9 9 9 47 Five
1 l 1 2 2 2 9 18 l 2 9 9 9 9 9 48
1 1 2 2 2 2 9 l9 2 2 9 9 9 9 9 49
l 2 2 2 2 2 9 20
2 2 2 2 2 2 9 21 l 9 9 9 9 9 9 55 Six
2 9 9 9 9 9 9 56

l l l 1 l 9 9 23 Two
1 l l l 2 9 9 24 9 9 9 9 9 9 9 63 Seven
1 l l 2 2 9 9 25
1 1 2 2 2 9 9 26
'l 2 2 2 2 9 9 27
22 2 2 2 2 9 9 28

135

APPENDIX A-2

CALCULATIONS USED IN DEVELOPING THE "PFC SCORE"

Cut Points and Weights

 

 

 

 

 

Weight: 2 Weight: 1 Weight:

% % %
Protein (P) §_9.9 10.0-25.0 3_25.l
Fat (F) 519.9 20.0-35.0 3_35.l
Carbohydrates (C) _:70.1 70.0-40.0 §_40.l

Patterns

.3 F__ C_ Sggrg_ _2 "Meaning"
1 l l 1 All three adequate

Two adequate, one inadequate when:
l 1 2 2 P & F adequate, C inadequate
l 2 l 3 P 8 C adequate, F inadequate
2 l l 4 F & C adequate, P inadequate

One adequate, two inadequate when:
l 2 2 5 P adequate, F & C inadequate
2 l 2 6 F adequate, P & C inadequate
2 2 l 7 C adequate, P & F inadequate

2 2 2 8 All three inadequate

\JOWUT-DwN—J
o o o o o o o

136

APPENDIX A-3

CROCETTI 32 FOOD GROUPS

BEVERAGES

Milk

Fruit and Vegetable Juices
Coffee and Tea

Soft Drinks

Fruitades and Drinks
Alcoholic Beverages

OTHER FOODS

Soup

Eggs

Cheese

Meats

Meat Mixtures

Poultry

Poultry Mixtures

Fish, Shellfish, Seafood

Fish, Shellfish, Seafood Mixtures
Legumes

Plain Starchy Side and Main Dishes
Vegetables

Starch, Vegetable and Protein Mixtures
Bread

Hot and Cold Cereals

Fruits

Frozen Desserts

Non-frozen, "Spoon" Desserts

Cakes, Cookies, Pies, Pastries, etc.
"Snacky" Foods

Candies, Sweets, Chocolates
Garnishes

Butter and Margarine

Sugar Garnishes

Nut Butters

Nutrient Supplements, Meal Extenders, "Health
Foods"

 

137

APPENDIX A-4
SELECTION OF VARIABLES TO BE INCLUDED IN PHASE II

Glossary of Indicators:

A cleared data tape was copied from Crocetti and Guthrie (1982)
secondary analysis of the NFCS. It contained data on variables that were
actually coded as they had been by USDA on the household and individual
questionnaires. It also contained variables that were constructed for
usage as in Phase I. (See Chapter 111., Section 2.). In this appendix,
a listing is presented which identifies variables as they were found on
the data tape and/or how they were combined or categorized for usage
in Phase II. The list serves as a glossary of indicators for the model.
It is organized in four categories based on the sets of variables used
in estimating the model: (1) problem affecting intake; (2) demographic
characteristics; (3) personal and food related behaviors; and (4) nutrient
intake. The list only includes those variables that were used in the
model.

Unit of Measure or Discussion of Variable
Variable Name Categories of Variable Construction

 

(1) Problem Group:

 

Dependent Variable Yes - respondents who
identified any non-
medical or medical
factor

Problem Yes (1)

No Problem No (2) No - respondents who
identified no
problems or
factors affecting
intake

138

 

 

 

Unit of Measure or Discussion of Variable
Variable Name Categories of Variable Construction
(2) Demographic
Characteristics-
Set of independent
variables includes:
XD Age Number of years Calculated from date of
l birth given on questionnaire
XD Sex Male (1) *
2 Female (2)
XD Race White (1) *
3 Black (2)
XD Education of Number of Years *
4 Male Head of
Household
XD Education of Number of Years *

5 Female Head
of Household

XD Poverty Poverty index *
6 (U.S. Bureau of
the Census, 1971)
XD City (1) Central City * Dummy coded
7 Characteristic (2) Suburban
and (3) Non-Metro (XD - City C -
7 CentraI City
xD xD - City 02 -
8 8 Suburban)

139

Unit of Measure or Discussion of Variable

 

 

Variable Name Categories of Variable ' Construction
(3) Nutrient Intake -
Set of independent
variables includes:
3 Day Average
XN % Calories percent +
1 from Fat
3 Day Average
XN % CBIOFTES percent +
2 from Carbohydrates
XN 3 Day Average percent T
3 % RDA Protein
X 3 Day Average
N4 % RDA Calcium Perce"t +
X 3 Day Average
N5 % RDA Iron Percent +
X 3 Day Average
N6 % RDA Magnesium percent +
X 3 Day Average
N7 % RDA Phosphorus Percent +
Log of Log
XN 3 Day Average percent Transformation of
8 % RDA Vitamin A calculated value
XN 3 Day Average percent
9 % RDA Th1am1n +
X 3 Day Average
N10 % RDA Riboflavin percent +
X 3 Day Average
N11 % RDA Vitamin 86 percent +
X Log of 3 Day Average Log transformation
N12 % RDA Vitamin 812 percent of calculated value
X 3 Day Average
N13 % RDA Vitamin 0 percent +
X 3 Day Average
N14 % Recommended percent +

Energy Intake (REI)

11E} ”rat-rm“

Variable Name

(4) Personal and Food

 

Related Behaviors-

 

Set of independent
variables includes:

x
F1

Weight
Status

Milk
Food Group (#1)

Coffee/Tea
Food Group (#3)

Meat
Food Group (#10)

Legume
Food Group (#16)

Starchy Side and
Main Dish Food
Group (#17)
Vegetables (#18)

Bread
Food Group (#20)

Cereal
Food Group (#21)

Frozen Dessert
Food Group (#23)

Cakes, Cookies and

140

Unit of Measure or

Categories of Variable

percent
(1.00 equal within
normal range (+ .10
equal 10% over nor-
mal range).

# of times men-
tioned food group:
3 days total

Pies Food Group (#25) "

Butter and Margarine

Food Group (#29)

Discussion of Variable
Construction

Discussed in Chapter
III.

141

Unit of Measure or

Discussion of Variable

 

 

 

Variable Name Categories of Variable""‘ ‘Construction
(4) Personal and Food
Related Behaviors
(Continued)
XF Nut Butter # of times mentioned
13 Food Group (#31) food group: total of +
3 days
XF Ate Alone- (1) Yes Dummy coded from Ate
14 Some Meals Day 2 (2) No Alone Day 2 category:
ate alone >1 meal, but
less than all meals.
XF Number of Days # of Days Crocetti, A.F. and
15 Ate Breakfast Guthrie, H.A. (1981)
XF Number of Days " "
l6 Ate Lunch or
Brunch
XF Number of Days
17 Ate Dinner " "

* Identified as was coded on NFCS questionnaire or as USDA transformed
respondent's answers and coded on NFCS tapes provided to contractees.

+ Crocetti, A.F. and Guthrie, H.A.

Final Report:

A secondary analysis

of the NFCS to Study Food Consumption Patterns of the U.S.

Washington, D.C.:

53-32U4-9-192).

U.S.D.A., 1982 (U.S.D.A.-CNC Contract No. 53-

142

APPENDIX A-5
"COMPOSITE INDEX" CATEGORIES

 

 

 

'PFC Index'
Marginality One Two or More
Index- 7* All "okgyf Outside Range Outside Ranges
NONE A B C
ONE 0 E F
TWO G H I
THREE-
SEVEN J K L

 

* Marginality Index represents the number of nutrients ingested at less
than or equal to 59.9 percent of the RDA. The seven nutrients in
Marginality Index: calcium, iron, magnesium, vitamin A, vitamin B6,
vitamin 812, and vitamin C.

>
1

All macronutrients within ranges weight '1' and no micronutrients
less than or equal to 59.9% RDA.

One macronutrient outside of specified range and no micronutrient
less than or equal to 59.9% RDA.

Two or more macronutrients outside of specified ranges and no micro-
nutrient less than or equal to 59.9% RDA.

All macronutrients within ranges weight '1' and one micronutrient
less than or equal to 59.9% RDA.

One macronutrient outside of specified range and one micronutrient
less than or equal to 59.9% RDA.

Two or more macronutrients outside of specified ranges and one
micronutrient less than or equal to 59.9% RDA.

All macronutrients within ranges weight '1' and two micronutrients
less than or equal to 59.9% RDA.

One macronutrient outside of specified range and two micronutrients
less than or equal to 59.9% RDA.

Two or more macronutrients outside of Specified ranges and two
micronutrients less than or equal to 59.9% RDA.

All macronutrients within ranges weight '1' and three-seven micro-
nutrients less than or equal to 59.9% RDA.

One macronutrient outside of specified range and three-seven micro-
nutrients less than or equal to 59.9% RDA.

Two or more macronutrients outside of specified ranges and three—
seven micronutrients less than or equal to 59.9% RDA.

APPENDIX B
PHASE II:

PRINCIPAL COMPONENT ANALYSES
AND CORRELATION TABLES

 

‘A‘_- .4—1- “Ad “H
F... - h.

143

.mcouoowucw mpaowco> pcmucmamvcw m>wuumammc mg» we some
Low mmzpo> mcvmmws o: saw: mNAm mpqsom on» pumpwmc m.z mg» .Amo¢.~uzv
HH omega cw cowumngoa auzpm one we mFQEomnam Eoucoc xofi on“ yo wpm: oz «

 

acacosoowaOOn. opacaoOno o nucpct sou-«aunt: no pica-lxu—aa

. o. «.0 cc... -~c. ._nap. aono~.u ¢~8.c.t pan.:.o ~u...u
“an.” ”acmo.— u«o.a.o m..oc.t ~ p. . .«.~.. panoc.o -e.:. .3»...
we . acqc.o ecoc.. «wear. a e._. rug ~.- ¢.A.c. ”mm.c. ....eo“
~.w. ..o .o S::.. coc.. ~«um. aw.v..o .~ ec.o :..o ..:e
'0'. s . . -¢e.. .~A. . coc.... syn—o.o .-o.. nec~ .u cyan.
«e~.o .m«~u. -n~.t “w.~_.o ..o.o.- ca:.. oa;.:.o ¢~:,g.o .6.-
o—c.o -ou.t ¢~.c. .09.- -~o.. oao.c.o woaaxn. ans. ..a
’39.. m~o. . n~o.c. noo.:.o sneuc.u o~c,c.c .cm ccc.— we.
a...» .u-.u ppuaaoc .uaou uuza. .... ..u .ae

 

 

Aeflmmuzv mowpmwcmuoocogu uwgaocmoEmo mo weepmuwucH "chuoz covumpmccou

.Hum u4m<H

TABLE B-2.

Indicators of Nutrient Quality (N=1,177*)

Correlation Matrix

 

 

CllO‘IO IIAVIO I0l(ll( .003! OOIIIS IOO'IOS' tIOIUIVO IOAVIII IIOUIOO

(ILOIIV

KOnOOQNOCOOfiflo

.”¢-c*~OCfl°fl-
¢-—~°~aao~°.¢.
UACOJFQCOOd‘OflC
--~"O"°fi.
QFWW-IOFOOO
0000000000000
0 o

doovn~¢6~0-Qn-
oacs~¢~0ndao¢0

c O ."" "vafifl’.
0000000000000.
.

C'WC'C'Mwfl.

Cpqrfio*~CVCr.0
vfiuuuoac~—-ndo
cnvv~0n00~aoon
¢OJJ¢3-¢aocoa
cc-oc~o¢ncer~
00000000000000
0 p

~Q~dcrnr~0¢0-

no~ooon~onon~c

~Q~Pu~m~OOJ¢4O

Odvacvpmdoo~n¢

corveohflvcrd~n
00000000000000
0 u

OdoOvopcbeOOOO

U'CO‘CFOO'mwsp
040083fl0'xaafis

00F0¢porn0¢onn
NOJOOfiN$GOO~OC
rasflcrzfimmbcoo
'OocaoooPOpgep
CGFOOOPO'CO"°
0000000000000.
.9000 I O I

anco~o¢0¢~nupo
ascon—~ooo~ao~
On~cbcc~w0a~on
O'~000¢V—030n~

cl O .000 O
0‘

b. U n... U

CSOd U0--. ~-

-UICJ‘S’I-J--

0009-001-0810

aa<¢qc 010030-

HUI-CIIJCCIJCC

lIOADZ I'll!!!

144

3’-O~OOF.°¢DOO

vhflcfin
00000000000000

oQOUTDODS-ODS)

000000003

OEIIIIIUIII 0' (IIIILIIIOI IA'III '

 

The N's

2,406).

N = Half of the 10% random subsample of the study population in Phase II (N

independent

1V6

ing values for each of the respect

th no miss

126 W1

reflect the sample 5

variable indicators.

TABLE 8-3.

19194*)

Correlation Matrix: Indicators of Food Related Behaviors (N

 

 

I061! Inf?! Iocrio Ionrio 0060'? IIG'II IIIP20 IIGPII III!!!

UISIIV

“flfiwvo oomovo~ov
G'GCOCOOO
0.000000000000000
I I II '
vr0n~000900000~90
scw~ooawoc~0
flocmaocvcwococ-N
MOCM~V\OO'¢‘O"OO“
cri‘o¢°m¥u¢°©~ O
0 00 00000000

oao-ovoafioooowon“

00~0Mn31a~00—vc-0

~¢v~r00°050fl0 00-4“:

ecu-4000'! )nonV—on-n

CwC0cO-CODO0-c~-v

00000000000000000
'I

0¢~~r~cnocv~0c420
saroocoofi~oaapw00
'fi‘VOICOffKNK‘I‘N.
00-CAh3nan~400~~n
cc “axe,“ v-cc o—ccccc
00000000000000000

Jol‘fi'vé 0.3V

r0000¢~h0r¢¢¢nhn~
~nJ-ovnswooofi030-
0"4~~0F0~O~'~O-
‘- N‘“-~’~- ‘0 0’30
0 t>3CC' ‘JL.C'¢-'.Cf‘€
00000000000000000
Iv I II I I

hgdowgwfld‘N’O".
cow—~ccocsancm~oo
(‘9' If-d- ¢~~~’.'~~o-&m
Jo:’~04;~~‘Hsavm-'
‘Q'ccgfi0"~"°A'°
00000000000000000
I0- I I I

5fl0w99$“0“~°"
x‘swn¢~4~°oovo *0
Go (:1. x1 0'80\€J'\~0-0- 4
$82-90¢~—IQ~~—on
3049-9c9400453

n W
h co~ U¢~now cue
‘pn~~.‘ .¢O\Vl"

C
P000830 6k A‘Inuflw
”99333-3 JJJ-DDD
pOODCOO OOO‘CQ‘
I--------fl---d...

3
3&0
3":
GO-

145

000029 IDGHI (0'22 INSIIOI 00.5001 00'3003

IO‘IZS

q

I

9

6

9

II

0

2

0

I

0

I

I
V'c"

 

¢n¢a¢~cg¢ncuncwg~
NJ»04‘QO"O¢\¢ O".- -30
'o~cr0~rvr—n:~cg~
ca~aea~wooc 0:00:30
C'OOOcc'CO'O'COO'
00000000000000000
I I II I 'I

 

cascaupnvvai‘qsoon'.
c~ooo¢acwon-ac~0
pro~~manquhanhnns~l~
unofvocsnomonunONN
CNC'CLCNN‘P'OULCO
00000000000000000
9 I0-

krrh-OOOOWQ'NOC

~a~2~40ona~aoano 0

ocConcccccocccco
00000 00000000
I IIImII IIII'II

“NCN'CV‘U‘I‘C NOCWKN
.J‘~J 06-0“VNJM‘6
d‘Nn-OOONNI'n-v0‘Ot'pKfi'i
~~0 aa>0‘)003 304‘)"
90"! ccfc-HPPaC-TC'C
00000000000000000

ova~¢n~ccucocan0
0~oc~nwc0wnc~o~0~
0'50! “908””0-850' 0"».‘c

ana'~¢hn¢J;cno~d
prroO'ccocccOpga
0.000000000000000
I I -..

omov~0m00°°c~a000
o)o~onoo‘bno~~0~nfi-
orJvco~00rD~amww~
“O'Qfi-wnenoa¢~naa
amouooow—uaaocvru
0000000000 000000
I II c I I

n unn

cos scw~w0P coo
=-A--'v-Cv~a.~~n~t ( I
b050006000000~waa
“JJJJJJJJOJJJ-DDD
.0600600000660000
0.-------a---u00

02250005'0300’

oZZSIIIOI

DIIIWIIAI' OI (OIIILIIIOI RAVI-ll |

 

The N's

independent

in Phase II (N=2,406).

ion

Half of the 10% random subsample of the study populat

*' N :=

1V8

lues for each of the respect

'th no mi551ng va

1Z6 W1

reflect the sample 5

variable indicators.

 

TABLE B-3.

1,194*)

Correlation Matrix: Indicators of Food Related Behaviors (N

 

 

IDGII Iorrs Iocrio Ionrio IOGrll IIcrII IIIP?U I06!!! I06!!!

IISIII

antcooo0aooooe~na

"~‘-°”.°°‘O‘Q~.fl
Q'OCOCOOC

00000000000000000
I I II V I

nr0n~000900500N00

-~ooa mm
oous$oc¢0noo¢o~
ocmnmoaoaonoon
C‘OOOOOG'COO~GO
0000000000000 000
IIII I- I

C'COCO'CCDP"°~—.
00000000000000000
'0 I

0¢~~v~c~0¢vhv0000

‘CflNGC"CCOFCCCCC

m0000c~::rCCana~

N/‘D 0° 0” 00"} 3c-

9°80~~G:;:00r~0-

~0-—~)~F500,30

0 OGCCHQC.C’—L¢PC

00000000000000000
I- I II I

AdeQ"fimwah~wono
eon—Vccoc~>~ca~oo
(‘9‘ I'vd- {VQN’ANN'CW
JO¢w~OJ¢~~O~I~avm¢
tovccuo-flh~-uOApc
00000000000000000
I'III

’n0w00$“0'~~ O“.
:Osvnqhosfioovcdhd
00011 ‘1 "\CJ'\~-F{
60:2—uoosot9~~pon
nu 3vo-o¢oao«--a
0000000000000000

n

F Co~¢ovun0u '
"‘~~ . ‘& . “““
bttttttsautssumnu
“QJJJJJJJOJJJ-DD.
“GOOGOOOOQCOOCCC‘
‘b-------0‘---d...

145

I06I29 IIcrsI (0'22 00'5001 0013-02 00'3001

IOSIIS

QOKCCKCOO'CCNOFNO
0C

OOCOCOt'cOCCCGO'G
00000000000000000
II I I-

pug-v-cscg‘rcu: CO‘C’N
~Jaoaae—-0»0o «5-00
'O~crd~rvru0fflvfl~
0a~00>~-009030~¢O
evoccccoco.0ucoaw

00.000000000000
I I II I —I

 

cosoquwn~~(¢~oca~

o~ooocac¢0n-ao~o

-V‘ONNW'fl'fifiN'ae!‘

'nowc0~n¢~n¢ano~~

c~c—:cc~~¢wvogcco

00000000000000000
In

~""°°‘¢°M.P°‘°C
3‘00 0 ”'flM‘OV‘OfiOO
~H~quq¢pwmcngo

Q JAN: 0"- 0“

oncvsan~o0~¢Ocdn0
0~oo~bwu0wnc~o~0~
DIKVMOONhMfl-suf0\.\.\c
wcna—~0na¢¢;ca0~0
vavoO¢0OOCCCOPGO
0.000000000000000
I ’II

~Ow00000~n000
woow~~a~na-
J¢CQ~00¢D~¢¢¢¢~
"fi'~fldn3’d~naa
r-Guooc-POJ5-Cp'u
0000000000000000

fl PNU‘

OK ‘c-~ao' coo
=--——-—¢.~;~~n~g I ;
"L0fi‘k‘$éfitt~nflfl

000050600000
I.------.a---o¢0.

03230083'0300I

62250000I

IEIIIRIIIII OI (OIIILOIIOI IIIIII O

 

2,406). The N's
independent

1V6

N = Half of the 10% random subsample of the study population in Phase II (N

lues for each of the respect

'th no mi551ng va

TZE W1

reflect the sample 5

variable indicators.

 

146

.mpmm mpamwgm> ucmucmamucw
m>wuumammg mgu $0 :umm com mmapm> mcwmmwe 0: saw: mNPm mFaEmm as“ mpompvmg z msh
.Aooq.muzv HH ommsa cw cowum—zaoa xuaum any mo mFQEmmnsm sauce; xop mg» 0o 0pm: u z «

 

oz.- ..2... .33.. 330.0 :3. mos. — 00. a p ..- 0:;
a... o~oco «op—0. ooawc.o om:c.o can. 0 «5.0 « ~ u. up.
as”- ....x“ xx"- are. 23 as" 5;” u ..H. .....WE
22$. «:3. a: .0 «Jam. 2...... . naaw . mac. mm 2.. £3.
us . nooscoI 99. .I s .I c—u_woo won M.I ”pot no». ea-
ooaoooo Onsncol c'rac. n—nao. —~ooc. r: . ~a— 0.. 00a o.o nun
ooopo. vnn~coo ecu... paauc. vonnco v ..c.0 «0.0:.0 na¢ ..I .6.

. coped. h 00—00. 0 00-u¢~ a 00:00. 0 00.0.. n 03-00. d 00pu§u — coped.

 

 

A«Hmmuzv wgsuusgpm Louumm Aumsgoymcmgpv vmumuom
och ”muwgmwgmpumcmzu uwcamcmosmo $0 mcopmuwucH mzu :o mwmxpmc< acmcoqeou pmawucwgm

.eum m4m<p

TABLE B-S.

The Rotated (Transformed)

ty:

1

tors of Nutrient Qual

1C6

' the Ind

15 on

1 Component Analys

incipa
Factor Structure (N

Pr

l,l77*)

 

 

IICIOI I“

I IIC'ul 2 IICIOI 3 IIC'UI I IhtIOI 5 IICIOO 0 IICIOI 7 IICIOI I IOCIUI 9

IIIIuI

wondn‘hNfipV—sh
osono~waon~0~0
~oooo¢0~-aooo

'*¢~~‘~~fl~...'
pvvcޢh~wnpn~Q
:'n¢'°~‘°'fl~°“
“0,5060.“
SONC"¢OGNK'OO
00000000000000
II

ohawcnONQC‘b-
0fl00a00-~3.-fl~

"- F'-
00000000000000

(c-000p00'NHQV
nosa~ao~~3m~

00 0~M¢¢kynhnc
do :OO~~¢OOOfl~
OC'FC ~C0-
00000000000000

moflusficrnvagn
na~qpavovoofi

cuck¢fi¢c-&00fi.

anwcmwon~ocun
'o'dz>ovccr<¢n
0"»

50000000000500
nu~ we‘d—“pep.
00$ FCfiO
a» ~n0aa~000
oc~v0~0w~0~nc~

0000 0000000

igrc00'a¢.~00~
.~o000m~a'o~0
tonw~o€noanpvo
'fl~‘- .AOGNMO
OONP‘CP- .m""
00000000000000

soonrwb~a000v0

nqa~onszwopano

90-0-0 stun; (pH-00¢

“JOFnhrPOwOONn

oopocoacccap—e
000 0000000000
I I. I IIII I

.mosunanovvfid

{$1 c—nooo-
“’9 Oihoffl¢~
“as ans Ch

oCnovwoom-«wsN-nn

00
b0 0 “'0. U
0861 u --. -
.00 d0 DI‘ofi-
0:00-300—0300
4000001000000-
0000000000000-
HUG-JICJIJIJCC

IICIOI It IICIOI 5" IICIOI II

IICIOI II

147

KfiONKOflN~NNOCO
-0U00fl00.n$°¢.
C6fi80n~dumu~ru
nJ—0°—~g~np-flp
CO¢C¢O~UCOOCGD
00000000000000
II II I

cooro~0~eoevav

Mao ~0fib$°3dfln

SOOfNCNNHO'CHO

J— Wfi-‘fi‘ ‘p~—O.

PCC¢U¢C¢.GnCCCU
0

‘~.~O¢:.K~°OB‘

-c 00:
00000000000000
I I II

800°000dbfiw$n0

00
p. 0 ”~06 U
0890 00--0 «P
-u‘CuCI’I‘O“
0000.000—0000
JJ¢((0¢0¢000CC
00000000000800
00000000000400

 

The N

2,406).

N = Half of the 10% random subsample of the study population in Phase II (N

independent

1V8

lues for each of the respect

0

'th no mi551ng va

126 W1

reflects the sample 5

variable indicators.

Table B-6.

The Rotated

lOVS:

tors of Food Realted Behav

(0
0"
.P*
'tJ<T
CZCh
u—n—a
O
Q)FH
J: H
+4 I:
v
C:
CO)
L.
0323
w- +4
VIC)
:>,:3
F- S-
“344
CIUO
<
L.
+90
CI+J
Q’CJ
CICU
CJLL
CL
E54-\
(3‘13
L) Q)
E
F- 8-
NS CD
CL“—
w- U)
L’C:
C315
w- L-
LI—
Q_\’

 

 

IICVOI '0

IICVOI l IIC'O' 3 IICIOQ I Il(100 S IICYOI 0 IICIOI 7 IICVOI I IICVOI '

I

IICIOI

(OQCMMO.‘N¥W.O¢wwflvfi.
oomwnm005~vnu0000013
omen—'ofik‘fldh~~00
~no~woo~cnwoop00~
ocooco~occococcoo
000000000000000.

.O~’0OO.’O°O¢~B$C
"ONOOO’G M“

'Qa‘n'fla‘mwzem
MOO'H‘MF .30-m ~M
COPCOOOOOOOOOOCGO
000000000000.0.0000

Ni.‘ .QNQG CONN. G‘V‘
osniflifio-loC-OO-ROO'A
NOOOfivnfi'fiv-P00000nwl1
mauoov-anoonnuv:
CCOOCCCODOv¢ECOOI
0000000000000000
I I II

 

-
'UU'V'

~00~0000¢ufl:0~cocc

~o~o¢0n0MOQ—O—Om no
oooouoccaccccccoo
00000000000000000

N'0Cbocch0c9anon
no~orvo~0090w0000
0~moacr¢0~ooc~uov
pnwfi~o Afiasnonnoa
OvOCCCCCCvocCOCCO
00.00000000000000
I II I I

monocr0~~00500aoq
~~¢ooau~~¢amnaovn
FOOO'C'wIO0J‘w-‘Mrw-c-
~~cw~3nnv~000~aon

CWOUGGCCCOOO
000000 0000 00 00 000
I I II I I

CKOOC'V‘OGQ “OED”

nsovoowonooon—Vaa
Cococooococooocoo
0000 00 0

r-vk°¢ofinwh00w~n

mnOAMuu~000oouwcnw5

vaocoo-d'nwccm—O
UCGOOOONOOLCDCGDO
00000000000000000

I I II I

0"" o cam-mo duo-'0

n
b Chm~0cn~unow 0120
4.0-00—~0;V~fl~888
b000000000000~wnn
”000000000000-F”
POOR-09.600.000.01
30505555505054...

148

IICVOI 1d IIC'II 13 IICVOI II IACVOI 15 IICVOI I. III'OI I?

IICVOI 1‘

WPQ": ONO' .000-P
m~o~orm (WON‘Mu-NO
out-Tor puns NOQP‘NOOB no

0000000000000ncon

~roaoo~~~u~oso
oaooo~aa~>~00~oom
nmaooo~~0~mdqcccc

v-M "NN

o0osawoo-oc 8800
o~p~00mo~muo 000 00
o0- $005 pr.fin¢v~.~0.hp

some
000800300 nem~o~0
Ow'cwscnc {vacuum

0» -M
0- DW‘OW‘O' DCI:
00—00-0-0-0-N 00. a 0"»: :2
p000000000000~mnw
0000000000000-000
b6000000050000000
305550555n0550...

 

The N
independent

2,406).

N = Half of the random subsample of the study population in Phase II (N

1V8

ing values for each of the respect

155

'th no m'

128 W1

reflects the sample 5

variable indicators.

TABLE B-7.

Correlations Among the Indicators of Demographic Characteristics and Nutrient Quality (N=2406)

 

 

149

On~on~o¢nn
cooan¢wsn
Onhnmemqwn
0 0 0 0 0 0 0 0 0 0

xus

OOONO'NW‘
ommqnohnfloi
O'OQ'QNnMfl
0 0 0 0 0 0 0 0 0 0 0
'

xua

000080—010an
Onocnpwoooq~
oomo~eo~~onh

000000000000
p

xaa

OMONPnonIWP
9.064 {POI-PJ's:—
OPOOP—FPOOI-m

0000000000000
0- I I I I I

xuz

O~~m~nqonnwnon
C os~o charm—Own

le
o
I
1
o
o
9
3
3
1
1
1

PI I IIII I

8 ODQPMHGIO-DOQONOO
OOOMGDNDv-VINMP

“ OODOPOPDOODOOOO
00000000000000.

— II I

IN MHN'QOOONOGNBMIR

D OKMPNNPQV‘Q'INV‘NOM

K @400000006000000
0000000000000000

'I I I I I I
OPIBOOOPOJNCIO~I~OO
8 CmQPmPNMfiNKmov-DOO
X DDv-CDODOOvOOOOv-OO
00000000000000000
0-I

O'OcrvonOMOMOh-Nono
OOPNQNPQONMV‘NQ'F'
O'DOOODPDoPoOoo—PD

000000000000000000

"I II II

3 neoncnoo—n‘s Onoekemfi

OQV‘OOF'ONOPNoov—Nnmn
GMPrO-CCL‘CPDL‘OC.COC*°
00.000000000000000.
0- I I I I I I

OOfino-I-OOJMNOMNOmNM‘vl
ONT~IO~IVM~I~I~IPIVC¢D¢~0M~IA
OOPNF‘vaOOPO’FOODODDO
00000000000000.0000.
" I I I I I I I I I I I I I

x03

1v 0~QQMCmmc~chOpoeco«o
C3 JDofl—o~333-finw0~—o~~sp
x OOwODGOPONNQPhOPONNDP
000'00000000000000000
-1 I II IIIIIIIIIIII

... OQOONO~0MQIGOOO~PONIBOM

D henryed‘P-Dd‘deOt‘c-fimmflcwufl

K hOOODOOfiOpMo-o-v-OOC-NNA'O
000000000000000

. . . . . C .

" I I I I II I I I I I I I I I
MM.

"IN m wasmu—u—o—I a-n

xxxxxw xxmxxmxmxx

‘N14
1.000

x1113
o
2

ONOCQIF".
Q onecoo~a
x onmoomNN
Q 0 0 0 0 0 0 0

'

amonhq-NO
nommmsnw-O
ohcoooeek
S . . . 0 0 0 0 0 0

K 0-
Q
exact—II—u—o—a-d
222222222

000-000: ”(K

TABLE B-8.

2406)

(.1:

Correlations Among the Indicators of Demographic Characteristics and Food Related Behav1ors

 

 

xr5

xra

xra

xr2

xrl

X03

X02

150

oooscocnosons
00n0n~4NOKOoO
o—ooooooooo—p
0000.000000000
p

OOOODKOO‘O'OON

00000000000000
’ I II

OOOCQ'INOOO'OM'
OONPQQPO'VPQMW‘
ODDCNO'FO”°m°
........0......
P I

OnvNNQeflOOOOOOON
OnPNGOonnn;finnan
OPOODOOfiC-IVU-PDAJOO
00000000000000.0
PIII I

ooh~woo~nv0o~onn~
ooc~crm~s¢mss~~me
OOOOOOODOOOOOOOOO
0.0.0.0...0.0.0..
Fl II IIIIII

ONNnoQO—Qmopewssfi
macaque—peacikoooon
COOCOGOCDDOOOOOO’F
000000000000000000
wI I II II IIII

HODQNNOGN‘OMQ" {Inc-I0
cw’msOcrqspmkrNNPOO
cqooooooooonooo—ooo
00000000000 00.00

000
FI I I IIIIIII III

CeOMODOnenm~ewMO¢mso
O~~P~coowvn~m04Poo~o
Op—DOroooPoooooo—oo—

000.000.000.0000000.

PI I II II IIIII

Ohnnoo~hp~n00000—sopo
O'Nmfl'NODnOnNNOPOOP'N
DnOPOPQCPPOOQP—OQOJPF
0.000000000000000000.
P I I I II I IIII

ncnswo~n030n00~mkn0moo
oonnOonwfipnqu~on~ooP~
CFCCOCCCCPOCPCPCCC'CPF

0000000000000000000000

' I IIII I I III

D¢0~mcnhm0mshnoopqh¢pon
om~mcorcmrmomCm~NkPQPO~
DONNMPOO’DOPOOCF’CDOOOO
0000.000000000000 00000

0
PIII I II I IIIII III

OOO€Cw~ONPUGOkaOOQQm4¢N
c~00N~0930n~0—00Nonnqson
OOPOUOHOPOPOPOwOOOOOPOOO
0000000...0000 00000

00000
P II I I IIIIIIIIII III

om—smcmemnoeno~ponsncnnms
noemoc~wmnpm~nqsfimn¢no~oo
coon?fiOWPNflOOfiNfiOPPthOfiO

0000000000000000000000000

.P III I I II III I

6°00me
r ("I m'n‘ma fi—dfimfl

K 00"“ mxm‘xxxxxxxsoo‘xm

x x x
xr7 Xr8 xr9 xno xr11 xr12 xr13 xm r15 F16 r17

xrs

chemo——
Damcmoo
OOFOPOO

0000000
p

ONNONOmO
DONnOOOn
OFUOFOOQO
0000 0000

P I II

OQNOEOOQO
cochoommk
OCPCCCKOC
00000....
P I

COD-0'5““ (00$
Oo¢~PPmOON
OOOwQPOFDP

0000000000
PI

OMOOpcmctMo
ONOJJ~Mw~nW
OPOGOPODDCO
.0.0.000000
P II I

Downpoomnvne
Dome—roomann
OOOOOO'fi—fifin
0.00.0000...
w III II
o—umvmom
0‘ a-q—nu—c-t—l-fl
Lb“, 1..me
xxxxxxxxxxxx

TABLE 8-9.

2406)

Correlations Among the Indicators of Nutrient Quality and Food Related Behaviors (N

 

 

xuz XNJ xna xus ‘Ne ”n7 xna xN9 xN10 lel leZ xN13

le

151

onspnenNNOQveocnOoo
mpflmmmmpwc cm~~0~
Ono—POOONONPODPONOO

000000000000000.0000

cocon0008cc—0ncspnop
ooocnonOPMNOn—hcsmnn
Chance mappv-po-v-Oov-o-Oc
00000000000000000000
0- I I I I I

woomonconnoooomno—
OOPONONOONO cmmchcwpq
OOQOQflNPPPPFmPO-OQFPOQ
000000000000000000000
— I I I I

cooo0p00fi0~00w~00w~ooa
OANNNOJPPOJPnono-NNNn—a
OsbnOOOrv—OO—anNO-PFND—
0000000000000000000000
0- | I

owoneooonnonmcwo-ocnooo
DU‘OMGxCNCmNU‘V‘OCQQNV‘OU‘OO
osciJNSfiDNF—PnMOPmeO
00000000000000000000000
F I

ONOONOONNCNONOMDNNONDiON
DWOONMWwNmmev-osmnwcm
DcwdcneoF‘OOOOQC-npo-v-OONOO
000000000000000000000000
F I I I I

nacn—mmnmwnsnnnmmanfiv—fl
CLOCrOQCFLOOCv—KU‘FNNKOWFOQ
OanomMDqONDPPNpOwo-OOPQO

00.0000000000000000000000
'

o{GENOOOO'NNO~IOMCDNOONPPOC
NN~IOQ~I DOOwOOmQPulCPNOPOPOS
NGONOGV‘NOQCOPOMMN’NONOO
000000000000000000000.00.00

1.00

OQQMNOOflUNFNmQOMKPNOOPNP'
Orv-monN-nmooeeooomoevlerw-ormn
O«sumenonOFPN'P’NtVOPo-CDPOO

000000000000000000000000000
'

n~cmon¢50nwn~moekmficnoo-nonv-no
O‘Donnwnm'swsrumoneovonnOMOIsN—m
(248‘wkmmramCoPcCcc—mmpmprcrco
0000000000000000000000000000
0- I I I I

Ovnvmoow~o~o~oompwowmoo—snecc
CGNOch-NrerONNmowssqumomov-o
O«mum:o~~on~o~r~~0~~~pwpwmvoo
0000000000.000000000000000000
0- I I I I

OQM‘CQINCIB‘QNOUN'ODPV‘NQU‘OMQC‘O'N'
03009000”).-oa—mamnoov—o—oo0-fiorw-n
00-000POPOOPNOOPpMOCPDMOPOv-OPDO
00000000000000. 00.000.000.000

0
C-III I I II II III

mwommompsmmosmmoooepo—
{50-h.0Ommfln-Pfih-MM«fiMONNDJr-v-cflmv‘fifi
meOPmOnPG-PDOPnOPPONOOI-DOOOO
00.0000000000000000000000000000
PI I I I I I I I I I I I I I I
(Dc-0M- OWN
~63“; wwww~~———mwNG~—~m '—
22: zzzzzzanWmflumgM lb
X)“ mxuooaxto‘wxxxxxxxxxm x

TABLE B-9.(con't)

‘rr ‘ra ‘ro ‘r10 ‘r11 ‘r12

‘56

‘r1 1‘r2 ‘r3 ‘rn

'uld

1552

OCDMWDGNNWOO
CKO'Q’PU‘ON
.JOCKaoFCMwap
0000'D00000
0-I I

Iaannu>ocwuren
CONGOF’WVNE°£
OCN39FWNNDOCHD
00000000000
' I II

'0WNN‘ORO'

‘
.07!
-009.

6

I

7

7

P

g

I

3

o—«um-omkumoon
IDOQWNIQOO<MM¢ON
0"”N3OOCKNNDDO

000010000400000

canvo'M—oomsn
Ohmnuflv-C'JNCJNO
OOfi~n~PGOCNouQO

00000000000000

(Mknflhnnnodmunnofln-
Oh‘WV”~¢OO'WWNVPO
occc—c—ccp—cocc
000001I0001I00000
a» II II

anO—oposmso-cfvo
am~~or~o~~mpmcc~t~
O-Oo :o—n-~—v--~:o
0000000000000000
'III I I

Chum~okunno<mm00evma
O¢=v3¢up¢pa~nwfinn
CKMMDCCHDDCKNDOCMmDCKS
00000000000000000
.4 III III! II I

OlnmnNCMm>udw~umoosuw‘
ruuorwhnocusofimuaswuw—o
OCNMJNG—MwnOFwHMDOaiuh

00.000000000000000
"I

.' ChquOUMM~
“WNW—“*—
ZHAAHbehfiufihbhfifidfibu

xxxmnonunnxuxumuuonx

‘rnq l‘ns ‘rls

‘rna

LIST OF REFERENCES

153

REFERENCES

Adrian, J, Daniel, R. Impact of socioeconomic factors in consumption of
selected food nutrients in the United States. American Journal of
Agricultural Economics, 1976; 58(1):31—38.

ANAREM Systems Research Corporation. Secondary analysis of the 1977-78
national food consumption survey: A research proposal submitted to
the Institute of Human Nutrition. Washington, D.C.: U.S.
Department of Agriculture, 1980.

Aquwa, 0C. An investigathm1 of the nutritional implications of the
definitions and measurements of food consumption patterns for
developed and developing countries. Washington, D.C.: Howard
University, 1980. (Unpublished doctoral dissertation.)

Bass,A, Wakefield, L, Kolasa, K. Community nutrition and individual food
behavior. Minneapolis: Burgess Publishing Company, 1979.

Beaton, GH, Milner, BA, Corey, P et al. Sources of variance in 24-hour
dietary recall data: Implications for nutrition study design and
interpretation. American Journal of Clinical Nutrition, 1979;
32:2546-2559.

Blalock, HM. Theory construction: From verbal to mathematical
formulations. Englewood Cliffs: Prentice-Hall, Inc., 1969.

Burk,MC, Pao, EM. Methodology for large-scale surveys of household and
individual diets. Nashington, D.C.: U.S. Government. Printing
Office, 1976. (Home Economics Research Report No. 40.)

Caliendo, MA. Nutrition and the world food crisis. New York: Macmillan
Publishers, 1980.

Cambridge Computer Associates, Inc. The crosstabs system (Release 400
ed.) Cambridge: Cambridge Computer Association, 1967. (For
exclusive use by City University of New York Computer Center.)

Carter, RL, Sharbough, CO, Stapell, CA. Reliability and validity of the
24-hour recall, Journal of American Dietetic Association, 1981;
79:542-546.

(Ilark, F3 Recent food consumption surveys and their uses. Federation
Proceedings, 1974; 33(11):2270-2274.

 

"il

154

Committee on Food Habits. Manual for the study of food habits.
Washington, D.C.: National Research Council, NAS, 1945 (NCR
Bulletin #111).

Concern, Inc. Overfed and Undernourished? Points of view: A nutrition
report, 1979; 1(1):1-4.

Crocetti, AF, Guthrie HA. Report on Step II for the study of food
consumption patterns in the United States., Washington, D.C.: U.S.
Department of Agriculture, 1980. (USDA, CNC Contract No. 53-32U4-9-
192.)

Crocetti, AF, Guthrie, HA. Report on Step III for the study of food
consumption patterns in the United States. Washington, D.C.: U.S.
Department of Agriculture, 1981(a) (USDA, CNC Contract No. 53-32U4-
9‘192)0

Crocetti, AF, Guthrie, HA. Report #5 for the study of food consumption
patterns of the United States. Washington D.C.: U.S. Department of
Agriculture, 1981(b) (USDA, CNC Contract No. 53-32U4-9-192).

Crocetti, AF, Guthrie, HA. Food consumption patterns and nutritional
quality of United States diets: A preliminary report. Food
Technology, 1981(c); 35(9):40-49.

Crocetti, AF, Guthrie, HA. Final Report: Food consumption patterns and
nutritional quality of United States diets. Washington, D.C.: U.S.
Department of Agriculture (1982, in press) (USDA, CNC Contract No.
53-32U4-9-192).

Cronin, FJ. Personal Communication. Washington, D.C.: U.S. Department
of Agriculture, 1980.

Cronin, FJ, Krebs-Smith, SM, Wyse, BW, Light, L. Characterizing food
usage patterns by demographic variables. Washington, DC: U.S.
Department of Agriculture, (1981, in press.).

Dixon, WJ, Brom, MB. BMDP biomedical computer programs, 2nd ed. Los
Angeles: UCLA, 1979.

Dodds, JM. The handy five food guide. Journal of Nutrition Education,
1981; 13(2):50-52.

Dwyer, J. Executive summary: National conference on nutrition education:
Directions for the 1980's. Journal of Nutrion Education, 1980;
12(5):79-88.

Eppright, E, Fox, H, Fryer, B, Lamkin, G, Vivian, V. The North Central
regional study of diets of preschool children, nutritional knowledge
and attitudes of mothers. Journal of Home Economics, 1970; 61:407-
413.

Fink,EL. Personal Communication. East Lansing, Michigan: Michigan State
University, 1979.

155

Friend, B, Marston, R. Nutritional review. Hyattsville, Maryland:
Consumer and Epod Economics Institute, ARS, 1975 (National Food
Situation).

Green, PE, Carroll, DJ. Mathematical tools for applied multivariate
analysis. New York: Academic Press, 1976.

Guthrie, HA, Scheer, JC. Nutritional adequacy of self-selected diets that
satisfy the four food groups guide. Journal of Nutrition Education,
1981; 13(2):46-49.

Hanushek, EA, Jackson, JE. Statistical methods for social scientists.
New York: Academic Press, 1977.

Hegsted, DM. What Americans are eating now: Preliminary results of USDA
Food Consumption survey. Paper presented at: Washington, D.C.:
Third Conference on nutrition and the American food system, 1979.

Hill,MM, Cleveland, M. Food guides: Their development and use.
Nutrition Program News, 1970; 1-5.

Hyman, MM et al. Drinkers, drinking and alcohol-related mortality and
hospitalizations: A statistical compendium. New Brunswick:
Rutgers University, 1980.

Jenkins, M. Dietary relationships existing between macronutrient and
micronutrient intakes. University Park: Pennsylvania State
University, 1982 (Unpublished doctoral dissertation).

Karvetti, RL, Knuts, LR. Agreement between dietary interviews. Journal
of American Dietetic Association, 1981; 79:654-660.

Knapp, TR. Canonical correlation analysis: A general parametric
significance testing system. Psychology Bulletin, 1978; 85(2):410-
416.

Kolasa, KM, Lackey, CJ, Slonim, AS. Factors affecting food selection:
The roles of food science and of nutrition. Washington, D.C.:
National Academy of Science (In press, 1982).

Lachance, PA. A suggestion on food guides and dietary guidelines.
Journal of Nutrition Education, 1981; 13(2):56.

Learner, RM, Kivett, VR. Discriminators of perceived dietary adequacy
among the rural elderly. Journal of American Dietetic Association,
1981; 78:330-337.

Light, L. Changes in the food guide. Washington, D.C.: National
Extension Workshop on Nutrition and Food Preservation, 1977.

Light, L, Cronin, FJ. Food guidance revisted. Journal of Nutrition
Education, 1981; 13(2):57-62.

Luruj,LA and Burk, MC. A multidisciplinary analysis of children's food
consumption behavior. St. Paul: lhfiversity of Minnesota, 1969
(Agriculture Experiment Station Bulletin #265).

 

156

Marr,JW. Individual dietary surveys: purposes and methods. World Review
of Nutrition and Dietetics, 1971; 13:105-164.

Marston, R, Page, L. Nutrient content of the national food supply.
National Food Review (USDA), 1978:41.

McLaughlin, MS. Discriminant analysis in communication research. In:
Monge, PR, Cappella, JN, eds. Multivariate techniques in human
communication research. New York: Academic Press, 1980:175-204.

Mead,M. Food habit research: Problems of the 1960's. Washington, D.C.:
National Research Council, NAS, 1964 (NRC Bulletin #1225).

Monge, PR. Multivariate multiple regression. In: Monge, PR, Cappella,
JN, eds. Multivariate techniques in human communication research.
New York: Academic Press, 1980:11-56.

Murray, J. Household food consumption surveys of the United States
Department of Agriculture: A historical review of data collection
methods, 1894-1969. Hyattsville, Maryland: Consumer and Food
Research Division, ARS, USDA, 1975.

National Academy of Sciences, NRC, Food and Nutrition Board: Recommended
dietary allowances, 9th edition, Washington, D.C.: National Academy
of Sciences, 1980.

National Center ft”: Health Statistics. Health-United States 1976-77,
Hyattsville, Maryland: U.S. Department of Health, Education, and
Welfare, 1977 (DHEW Publication No. (HRA) 77-1232).

National Center for Health Statistics” Health-United States, 1978.
Hyattsville, Maryland: U.S. Department cfl’ Health, Education and
Welfare (DHEW Publication No. (PHS) 78-1232).

National Research Council, NAS, Committee on Food Habits. Manual for the
study of food habits. Washington, D.C.: National Research Council
NAS, 1945 (Bulletin of the National Research Council no. 111).

Neter, J, Wasserman, W. Applied linear statistical models. Homewood,
Illinois: Richard D. Irwin, Inc., 1974.

Nie, NH, Hull, GH, Jenkins, JG, Steinbrenner, K, Brent, DH. SPSS:
Statisical package for the social sciences, 2nd edition, New York:
McGraw-Hill, 1975.

Olson, CM, Gillespie, AH. Highlights of the discussions: Workshop on
nutrition education research applying principles from the behavioral
sciences. Journal of Nutrition Education, 1980; 13(1):576-579.

Page,L, Friend, 8. The changing United States diet. Biological Science,
1976; 28(3):196.

Pao, EM. Food and nutrient intakes of individuals within families in the
south. Columbus: Ohio State University, 1977 (Unpublished doctoral
dissertation).

157

Pao, EM. Changes in American food consumption patterns and their
nutritional significance. Food Technology, 1981; 35(2):43-53.

Pennington, JT. Dietary nutrient guide» Westport, Connecticut: AVI
Publishing Company, 1976.

Pennington, JAT. Considerations for a new food guide. Journal of
Nutrition Education, 1981; 13(2):53-55.

Pirie, P, Jacobs, 0, Jeffrey, R, Hannan, P. Distortion in self-reported
height and weight data. Journal of American Dietetic Association,
1981; 78:601-606.

Sanjur, D. Social and cultural perspectives 'hi nutrition. Englewood
Cliffs: Prentice-Hall, 1982.

Sims,LS, Paolucci, B, Morris, PM. A theoretical model for the study of
nutritional status. Ecology of Food and Nutrition, 1972; 1(3):197-
205.

Sims,L. Personal Communication, University Park, Pennsylvania:
Pennsylvania State University, 1980.

Stoyanoff, NF. An examination of the relationship between social
structural factors, complexity of the information environment,
individuals' processing style, and cognitive structure. East
Lansing: Michigan State University, 1981 (Unpublished doctoral
dissertation).

Stunkard, AJ, Albaum, JM. The accuracy of self-reported weights.
American Journal of Clinical Nutrition, 1981; 34:1593-1599.

Stunkard, AJ, Waxman, M. Accuracy of self-reports of food intake.
Journal of American Dietetic Association, 1981; 79:547-551.

Todhunter, EN. Lifetyle and nutrient intake in the elderly; In: Winick,
M, ed. Nutrition on Aging, New York: John Wiley, 1976.

United States Bureau of the Census. Poverty increases by l.2 million in
1970. Washington, D.C.: U.S. Government Printing Office, 1971
(Current Population Reports Series P-60, No. 77).

United States Department of Agriculture, Consumer and Food Economics
Institute. Review of commentaries and critiques on the subject of
food selection guidance tools. Washington, D.C.: USDA, 1976.
(Background paper for 1976 food selection guidance discussions.)

Lhiited States Department's of Agriculture and of Health, Education and
Welfare. Nutrition and your health: Dietary guidelines for
Americans. Washington, D.C.: U.S. Government Printing Office,
1980.

158

United States Department of Health, Education and Welfare, United States
Department of Agriculture, White House Office of Science and
Technology Policy, Federal Trade Commission, Society for Nutrition
Education. National conference on nutrition education: Directions
for the 1980's. 1980; 12(2) Supplement 1.

United States Department of Labor, Employment and Training
Administration. Dictionary of occupational titles, 4th edition.
Washington, D.C.: U.S. Government Printing Office, 1977.

United States General Accounting Office. Nationwide food consumption
survey: Need for improvement and expansion. Washington, D.C.:
General Accounting Office, 1977 (CEO-77-56).

United States Senate Select Committee on Nutrition and Human Needs.
Dietary goals for the United States, lst ed. Washington, D.C.: U.S.
Government Printing Office, 1977(a).

United States Senate Select Committee on Nutrition and Human Needs.
Dietary goals for the United States, 2nd ed. Washington, D.C.: U.S.
Government Printing Office, 1977(b).

United States Surgeon General. Healthy people: Report on health
promotion and disease prevention. Washington, D.C.: Government
Printing Office, 1979 (DHEW Publication No. (PHS) 79-55071).

Woelfel, J, Fink, EL. The measurement of communication processes:
Galileo theory and method. New York: Academic Press, 1980.

Woelfel, JC, Woelfel, JD, Woelfel, ML. Standardized versus
unstandardized data matrices: Which type is more appropriate for
factor analysis. Quality and Quantity, 1980; 14(6):699-710.

Yetley, E. A causal model analysis of food behavior. Ames: Iowa State
University, 1974 (Unpublished doctoral dissertation).