A

 

\

 

 

 

 

 

*f‘eSIQS

l//llil"ill7lllilifl7lllllm7i

3 1293 01055 022

This is to certify that the

dissertation entitled

EFFECTS OF DENSE PUBESCENCE ON
QUANTITATIVE TRAITS IN SOYBEAN
presented by

Ram Pratap Sah

has been accepted towards fulfillment
of the requirements for

Ph.D. degreein Plant Breeding & Genetics

 

 

,/

Major professor

r. James D. Kelly
Date February 27, 1992

 

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

 

 

 

LIBRARY
Michigan State
1 University 1

L

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES rotmn on or before ddo duo.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

V—ll l

MSU is An Affinnutivo Action/Equal Opportunity institution
cmmS-pt

 

EFFECTS OF DENSE PUBESCENCE ON
QUANTITATIVE TRAITS IN SOYBEAN

By

Ram Pratap Sah

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Crop and Soil Sciences
(Plant Breeding and Genetics Program)

1992

v‘

ABSTRACT

EFFECTS OF DENSE PUBESCENCE ON QUANTITATIVE
TRAITS IN SOYBEAN
By
Ram Pratap Sah

Pubescence density in soybean has previously been found to affect plant vigor and
resistance to insects and drought. There is inadequate information on how they affect
quantitative traits and their genetic variance. The objective of this research was to study the
effect of dense pubescence on the quantitative traits of economic importance, their genetic
variance, heritability, and correlation coefficients, and to discuss implications of findings to
soybean breeding.

Wells II with normal pubescence and Harosoy with dense pubescence were crossed and
progenies for normal and dense pubescence types were developed in a nested design, and
were evaluated for two years in F6 and F7 generations at two locations in Michigan. The
genotypic differences were significant for all traits. Trait means were affected by
pubescence density, year, location and their interactions. Pubescence density affected all
traits except seed size and protein. The interaction effect of year x location x pubescence
density revealed that effects of pubescence density were not persistent for most traits except
height. Pubescence density did not exhibit main effects on any trait, but interaction effects
were present with all traits indicating situations where pubescence density could be useful
in breeding.

In general, the variances due to pubescence density and its interactions were low compared
with additive and additive x additive variances for all the traits except height. Pubescence
density did not influence heritability. Heritability estimates were high for maturity, height
and lodging (>0.50); moderate for seed size and protein (0.35-0.49); and low for yield and
oil (0.24-0.34). The correlation coefficients were affected by year, location, pubescence
density and their interactions. Certain correlations, like yield with seed size and protein;
maturity with height and lodging; and height with lodging were strongly positive and stable
across factors. Protein and oil exhibited a strong negative correlation. Stepwise regression
was used to select variables for predicting yield, protein and oil. The implications of these
findings to soybean breeding for quantitative traits are discussed.

DEDICATED TO

. MY GRANDMOTHER

ACKNOWLEDGMENTS

I wish to express my sincere thanks and gratitude to my major professor, Dr. James D.
Kelly for his encouragement, advice and help. To my other committee members, Drs. Dale
D. Harpstead, Charles Cress and Amy Iezzoni, I extend my obligation and appreciation for
their valuable comments and critiques. Also, I feel indebted to Dr. Thomas G. Isleib, my
former major professor, who initiated and guided this research. I appreciate his help on
analytical procedures even after he moved to N. C. State University.

I am obliged to the National Agricultural Research Council (NARC) and to ARPP/U SAID,
Nepal for providing the opportunity and financial support to accomplish this study.
Financial support received from the Department of Crop and Soil Sciences, Michigan State
University, is also acknowledged.

To all the friends in soybean and drybean programs of this Department, I offer my
appreciation for their help. I thank Ms Betsy Bricker for her advice in final document
printing. I take this opportunity to thank Mr. Bharat P. Upadhyay and his family for
providing friendship and support in joys and agonies. Last but not least, I am thankful to
my wife, Shanti, and to my sons, Shashi, Sanjeev and Sujeet, who have always been very

inspiring and supportive.

iv

TABLE OF CONTENTS

Page
LIST OF TABLES ix
LIST OF FIGURES AND ABBREVIATIONS xiii
1 INTRODUCTION 1
2 LITERATURE REVIEW 6
2.1 Morphology of Pubescence 6
2.2 Inheritance of Pubescence 7
2.2.1 Pubescence Types 7
2.2.2 Pubescence Color 9
2.3 Effects of Pubescence on Insect Resistance 11
2.4 Effects of Pubescence on Agronomic Traits 12
2.5 Variance Components 16
2.6 Heterosis 19
2.7 Heritability 20
2.8 Correlation Among Traits 22
3 MATERIALS AND METHODS V 25
3.1 Development of Experimental Material 25
3.2 Field Evaluation ‘ 26
3.2.1 Experimental Design and Procedure 26
3.2.2 Data Recorded 28
3.3 Statistical Analysis 29

3.3.1 Means Comparison 29

4 RESULTS

3.3.2 Analysis of Variance
3.3.2.1 General Model ANOVA
3.3.2.2 Nested Model ANOVA
3.3.2.3 Correction for F-test
3.3.3 Estimation of Genetic Components of Variances
3.3.4 Least Square Analysis
3.3.5 Estimation of Heritability

3.3.6 Estimation of Correlation and Regression Coefficients

4.1 Comparison of Treatment Means

4.2 Analysis of Variance

4.2.1 General Model ANOVA
4.2.2 Nested Model ANOVA

4.3 Estimation of Components of Variance

4.3.1 General Model ANOVA

4.3.2 Nested Model ANOVA
4.3.2.1 Additive Model
4.3.2.2 Additive & Additive x Additive Model
4.3.2.3 Additive & Dominance Model

4.4 Estimates of Heritability and Gain from Selection

4.4.1 Additive Model

4.4.2 Additive & Additive x Additive Model
4.4.3 Additive and Dominance Model

4.4.4 General Model ANOVA

4.4.5 Parent - Offspring Regression

4.5 Estimates of Correlation Coefficients

4.5.1 Average Means Correlation
vi

30
30
3O
31
36
4O
43
46
48
48
55
55
57
61
61

9

69
72
72
72
75
77
77
77
77

4.5.2 Effects of Pubescence Density ,Year and Location on
Correlation
4.5.3 Effects of Interactions of Year, Location and
Pubescence Density on Correlation
4.6 Estimates of Regression Coefficients and Equations
5 DISCUSSION
5.1 Means, Simple Statistics and their Comparison
5.2 Analysis of Variance
5.3 Estimation of Components of Variance
5.3.1 General Model ANOVA
5.3.2 Nested Model A NOVA
5.3.2.1 Additive Model
5.3.2.2 Additive & Additive x Additive Model
5.3.2.3 Additive and Dominance Model
5.4 Estimation of Heritability and Gain from Selection
5.5 Estimation of Correlation Coefficients
5.6 Estimation of Regression Coefficients
6 SUMMARY AND CONCLUSIONS
BIBLIOGRAPHY
APPEN D ICE S
Appendix A .1 Overall Means of Genotypes across years and locations
Appendix B.l—B.7 Analysis of variance tables of the General model AN OVA
for Yield, Matmity, Height, Lodging, Seed size,Protein
and Oil
Appendix C. 1-C.7 Analysis of variance tables of the Nested model AN OVA
for Yield, Maturity, Height, Lodging, Seed size,Protein
and Oil

80

83
89
93
93
96
101
101
101
101
102
103
104
107
109
111
114

120

122

126

Appendix D.1-D.3 Estimates of correlation coefficients and their respective
significance among variables by year, location and

year x location 133

Table 1.
Table 2.
Table 3.

Table 4.

Table 5.

Table 6.

Table 7.

Table 8.

Table 9.

Table 10.

Table 11.

LIST OF TABLES

Expected mean squares (EMS) from the General Model A NOVA.
Expected mean squares (EMS) from the Nested Model ANOVA.
Estimation of Error MS (MSe) for the individual component line
in the Nested Design Model.

Coefficients for quadratic components for various covariances
for F2 and F3 derived lines in F6 and F7 generations.

Matrix of Coefficients of the Variances for the different
component lines in the Nested Design Model.

Overall mean, range, variance, Variance as percentage

of mean (Var %) and Coefficient of variance (C.V.) of the
characters.

Comparison of means and other statistics of the characters

by Year.

Comparison of means and other statistics of the characters

by Location.

Comparison of means and other statistics of the characters

by Pubescence Density.

Effects of Year, Location and Pubescence Density on the
significance of variable means using 'ITEST.
Effects of Year x Location on the significance of variable

means using TI'EST.

Page

32

33

35

39

41

49

50

51

53

54

54

Table 12

Table 13.

Table 14.

Table 15

Table 16.

Table 17.

Table 18.

Table 19.

Table 20.

Table 21.

Table 22.

Table 23.

Effects of Year x Location x Pub. Density on the
significance of variable means using 'ITEST.

Effects of Density x Year, Density x Location on the
significance of variable meansCI'I'EST).

Estimates of Mean Squares (MS) and F-test significance of
the characters in the General Model ANOVA .

Estimates of Mean Square (MS) and F-test significance of
the characters in the General Model AN OVA.

Estimates of Mean Squares (MS) and F-test significance in
the Nested Model AN OVA.

Estimates of Mean Squares (MS) and F-test significance in
the Nested Model AN OVA.

Estimates of components of variance from the General
Model AN OVA.

Estimates of components of variance in Additive (A)

- Weighted Model.

Estimates of components of variance in Additive,

Additive x Additive (A , A x A) ~Full Model.

Estimates of components of variance in Additive,

Additive x Additive (A, A x A) -Se1ected Model based on
Principal Component Analysis (PCA).

Estimates of components of variance in Additive x Dominance
(A x D)- Full Model.

Estimates of components of variance in Additive x Dominance
(A x D)- Selected Model based on Principal Component
Analysis (PCA).

56

56

58

59

62

63

65

67

68

70

71

Table 24

Table 25.

Table 26.

Table 27.

Table 28.

Table 29.

Table 30.

Table 31.

Table 32.

Table 33.

Table 34.

Estimates of Heritability, Gain from Selection (Gs) and Gs
as percentage of mean (Gs%) in Additive (A) Model.
Estimates of Heritability, Gain from Selection (Gs) and Gs
as percentage of mean (Gs%) in Additive &Additive x
Additive (A & A xA) Model.

Estimates of Heritability, Gain from Selection (Gs) and Gs
as percentage of mean (Gs%) in Additive & Dominance

(A & D) Model.

Estimates of Heritability, Gain from Selection (Gs) and Gs

as percentage of mean (Gs%) in General Model ANOVA.

Estimates of Heritability by Parent- Offspring Regression
(Regression of 1990 on 1989).

Estimates of correlation coefficients and their respective
significance among variables using average genotypic
means(AVE ).

Estimates of correlation coefficients and their respective

‘ significance among variables by Pubescence Density.

Estimates of correlation coefficients and their respective
significance among variables by Year x Pub. Density.

Estimates of correlation coefficients and their respective

significance among variables by Location x Pub. Density.

Estimates of correlation coefficients and their respective
significance among variables by Year xI.ocation x Pub.
Density.

Estimates of regression coefficients by General and

Stepwise Regression Analysis for Yield.

73

74

76

78

78

79

81

84

87

90

Table 35.

Table 36.

Estimates of regression coefficients by General and
Stepwise Regression Analysis for Protein.
Estimates of regression coefficients by General and

Stepwise Regression Analysis for Oil.

90

92

Figure 1.

ARPP
AVRDC
FAO
IITA

INT SOY
NGLIP
USAID

LIST OF FIGURES

Development of family structure.

LIST OF ABBREVIATIONS

- Agricultural Research and Production Project.

— Asian Vegetable Research and Development Center.
- Food and Agriculture Organization.

- International Institute of Tropical Agriculture.

- International Rice Research Institute.

- International Soybean Program.

- National Grain Legume Improvement Program.

- United States Agency for International Development.

Page
27

INTRODUCTION

Soybean (Glycine max L. Merr.) is an important legume as a source of human
food, animal feed, and materials for many indusuial uses. As a source of protein and oil, it
complements the conuibution of most cereal crops. Currently, the cr0p is grown on 58.2
million hectares with a production of 107.3 million tons worldwide (FAO, 1989). The
major producing countries are the USA, Brazil, China and Argentina accounting for nearly
87 % of the world's production. However, the importance of this crop as a rich source of
protein and oil, and role in soil improvement through nitrogen-fixation has recently been
recognized in developing countries including Nepal, where its area and production are
considerably increasing. Although soybean has been a traditional crop in Nepal, its
commercial cultivation is recently being realised with the development of new high yielding
varieties, production technologies and processing industries (NGLIP, 1988).

Soybean is considered to have originated in China, where this has been cultivated
for over 4,500 years (Probst and Judd, 1973) with its introduction into the USA in the
1800's. It was only after early 1900's, that the importance of this crop as a rich source of
protein and oil was recognized. Concerted efforts have been made since 1960's towards the
improvement of this crop worldwide (Caldwell,1973). The USA is the leading soybean
producer contributing about 41.2 and 48.8 percent to the world's area and production
respectively. Michigan is one of the important soybean producing states in the USA with an
area of 0.44 million hectares and a production of 1.05 million tons (A gri. Stat.,
USDA,1990). A

Considerable progress has been made in the past toward the genetic improvement
of soybean for increased yield and other agronomic traits. Currently, in addition to the
national programs, several international centers, INTSOY, IITA, IRRI and AVRDC have
a mandate to further improve soybean for yield, protein, oil, and other agronomic traits in

addition to resistance to major pest, diseases and stress factors. The significance of

2

germplasm in crop improvement as a source of elite genes is great. The USDA maintains

nearly 10,000 accessions of the world soybean germplasm which represents extensive
genetic diversity for the various morphological, physiological and isozyme characters
(Palmer and Kilen, 1987). One of the characteristic features of varieties grown in the US is
rather a dense covering of the erect hairs or the trichomes on the stem, leaf, calyx, and pod
(Carlson,1973). An individual trichome has a diameter of about 20-40 mm and a length of
about 0.5 -3.0 mm. However, considerable genetic variation in trichome size, form,
density, durability, and color occurs in the exotic germplasm (Bernard and Weiss, 1973).
Many of the pubescence variants have been genetically characterized, and most are
controlled by single dominant or recessive gene (Bernard, 1975; Bernard and Weiss,
1973). Bernard and Singh (1969) reported the existence of five independent genes
controlling the different pubescence types in soybeans. Pubescence density has been found
to increase plant vigor, reduce damage from insect (Empoasca spp.), and reduce
transpiration rate, providing this crop with better adaptation particularly under drought
(Hartung et al., 1980). Cooper and Waranyuwat (1985) observed that densely pubescent
indeterminate plants exhibited increased lodging. A differential response of the Pd-gene
(dense pubescence) in the determinate and indeterminate genotypes has been observed,
with an yield decrease in the later. However, densely pubescent genotypes tend to yield
slightly higher than those with normal pubescence (Hartwig and Edwards,1970; Singh et
al., 1971).

Most of the traits of economic importance in soybean are quantitau've in nature and
exhibit continuous variation. Currently used breeding methods for manipulating the
quantitative traits have their origin or rationale in quantitative genetics. The deveIOpment of
a more efficient breeding procedure is dependent upon a better understanding of the types
of genetic variance and gene action underlying the inheritance of the quantitative traits.
Johnson and Bernard (1963) reported that only additive component of genetic variance has

so far been exploited in soybean breeding for the quantitative traits such as yield.

3

Information on the various types of gene action and their relative importance in the
inheritance of traits of interest have been obtained in only a few populations of soybeans
(Brim, 1973). Hanson et al.(1967) observed additive x additive epistasis for seed yield,
maturity and percent seed mottling accounting for more than 50% of the total genetic
variance. Significant dominance variance has also been reported for economic traits such
as plant height, seed weight, and lodging (Crossiant and Tonie,1971). Since quantitative
traits are controlled by a large number of loci, linkage effects may become important in
interpreting these variances (Gates et al., 1960).

A clear understanding of the nature of gene action controlling quantitative traits is
crucial for the effective breeding of soybean. Also, the type and magnitude of gene action
are greatly influenced by the nature of the population, character studied, sample size
chosen, and the genetic models and analysis procedures used. Though additive variance is
the primary component for the genetic variance of quantitative traits in soybean, evidence of
non-additive variance for certain traits indicate the need for modification of the breeding
procedure for manipulation of such traits (Brim, 1973). Heterosis or inbreeding
depression may not adequately reflect non-additive gene action for self-fertilizing species
since dominance and epistasis deviations may be both positive and negative and will tend
to cancel out when averaged over loci. Many practical decisions in breeding program are
based on the magnitude of heritable variation. Estimates of heritability and their role in
predicting gains from selection have been done extensively in soybeans (Johnson and
Bernard, 1963). Most of these estimates are broad-sense heritability, which include
dominance and epistatic components of genetic variance. Since these components are not
fixed in the self-pollinated crops like soybeans, such estimates may not be very useful in
predicting gains from selection. However, heritability in narrow sense(h2) is more
important to the breeder because effectiveness of selection depends on the additive portion

of the variance in relation to the total variance (Falconer, 1960).

4

Correlation and regression estimates are also important to the breeders for use in
indirect selection , particularly when heritability of the economic traits is lower compared to
the associated traits. The correlated response studied to date do not appear to provide
useful selection criteria for increased yield. The results are inconsistent from cross to
cross, and over environments (Brim,1973). It appears that the usefulness of correlated
responses in selection for yield will depend on recognizing attributes other than those
usually measured and which are more closely related to physiological processes associated
with productivity (Brim,1973).

A knowledge of kind and amount of variability affecting important agronomic uaits
is essential for a successful breeding program (Croissant and Torrie, 1971). The isolation
of single genetic factors affecting quantitative traits is of great interest to plant breeders as
they can easily manipulated through breeding (Powell et al., 1985). Though there are some
reports exhibiting the effects of pubescence on quantitative traits in soybean, there is
inadequate information available to date on how dense pubescence affects genetic
components of variance of quantitative traits like yield, maturity, height, lodging, seed
size, protein and oil in soybeans; and how it (pubescence) affects correlation and
heritability estimates. Therefore, a more reliable information on these is essential to
drawing valid inferences and planning an effective breeding procedure for soybeans. Also,
these are crucial to the understanding of the underlying genetic phenomenon, used to
design an effective breeding program incorporating dense pubescence as a desirable trait for
the improvement of the quantitative traits like yield, maturity, height, lodging, seed size,
protein and oil in soybean. Breeding for the quantitative uaits is difficult and time taking
because of their polygenic inheritance. If the effects of pubescence density on these traits

are significant, it can be used for indirect selection in breeding for these quantitative traits.

Researgh Objectives :

1. To study the effects of dense pubescence on quantitative traits such as grain yield,
maturity, plant height, lodging, seed size, protein, and oil.

2. To estimate the genetic components of variance (additive, dominance, additive x
additive) for these quantitative u'aits under the different genetic models.

3. To estimate genotype x environment interactions affecting the components of genetic
variances.

4. To estimate narrow-sense and broad-sense heritability of these quantitative traits, and
gain from selection.

5. To determine correlation and regression coefficients of the quantitative traits under

study.

6. To discuss the significance of the findings in relation to breeding for quantitative

traits in soybean.

2 LITERATURE REVIEW

2.1 Morphology of Pubescence

Considerable variation exists among soybean genotypes with respect to size,
shape, durability, and distribution of plant hairs or trichomes. Singh et a1. (1971) studied
the morphology of five different pubescent types: normal, dense, sparse, curly, and
glabrous; and their effects on the agronomic traits in soybean. They developed near
isogenic lines through backcrossing in 'Harosoy' and 'Clark' for the different pubescent
types. They found that hairs of normal, dense and sparse types were morphologically
similar, each consisting of a very long (1-3 mm) cylindrical cell with one, two or three
basal cells. Hairs of curly pubescence were similar to normal type initially, but then
became flat, curled and tended to fall off. Glabrous plants had a hair stub made up of 1-7
nearly isometric cells. Puberulent plant hairs consisted of a single elongate (0.1 mm) apical
cell with 1-3 basal cells. They found significant difference for number of hairs in
different pubescence types. Normal and curly had 6-8 hairs lmm2 , sparse 2-3 hairs/mmz.
while dense had 3041 hairs/ mmZ. These number of hairs were also highly positively
correlated with leafhopper resistance. In addition, they found that dense pubescent plants
grew tallest followed in descending order by normal, sparse, curly, and glabrous plants.
Grain yields of lines with normal, dense, and sparse pubescence were similar, and superior
to the curly and glabrous lines. However, they did not find much difference for seed
weight, protein and oil composition of the seeds due to pubescence, except that the
extreme damage to the glabrous lines caused a reduction in seed size, and an increase in
seed protein in some environments.

Wolley (1964) found that hairs on the upper surface of mature 'Hawkeye' soybean
leaves were about 1 mm in length, and spaced about 1 mm apart, accounting for 10% of the

total leaf surface. Each hair consists of a long distal cell, 0.5-1.5 mm in length, which is

7

surrounded by a cushion of epidermal cells. The hairs are slanted slightly towards the tip
and edge of the leaflet. Mature hairs dry out and become air filled or flattened
(Dzikowski, 1937). In addition to these elongated uniceriate trichomes, small five-celled
club shaped u'ichomes are abundant on all young organs. These trichomes persists, but
gradually senesce in the mature leaves.

Franceschi and Giaquinta (1983) reported that the cuticle over the distal two cells
becomes distended, indicating that a secretory product accumulates beneath it. They
speculated from ultrastructural evidence that a volatile terpenoid compound is secreted

which helps to protect the developing leaflets against foraging insects.

2.2 Inheritance of Pubescence
2.2.1 Pubescence Types

Several workers have reported inheritance of pubescence in soybean. Nagai and
Saito (1923) reported a single gene difference between glabrous (P-) and pubescent (pp)
plants, with glabrous being dominant. They also gave evidence for a linkage of 'Pp'
(Pubescence gene) and 'Mm' loci with 18% cross-over measured on a large population.
'M' was a gene for black stripes (mottling) on a brown or buff seed coat, and the allelic 'm’
produced self brown or buff.

Owen (1927) also reported a single gene control of pubescence, with glabrous
being dominant. He used a glabrous mutant and a Japanese glabrous variety, crossed
with a pubescent type, and found that the inheritance was the same in both. He reported a
linkage of P 1p 1 with Rr ( black vs brown seed coat) in both the crosses, and suggested
that both glabrous lines carried the same gene for glabrousness. Woodworth and Veatch

(1929) obtained a glabrous mutant from Wentz (1926) and a dominant glabrous Japanese
variety from W.J. Morse of the USDA, and crossed them. The F1 was glabrous, the F2

segregated 13 glabrous : 3 pubescent plants, and the F3 segregation supported the

8

hypothesis of two unlinked gene pairs P 1p 1 and P2p2.. Due to the apparent but very short

pubescence of the p2 plants they were designated 'puberulent' type. Another distinct

pubescence type has been reported among Japanese varieties. Takahashi and Fukuyama
(1919) described that in addition to glabrous variety, two other varieties had hairs which
flattened, curled, and eventually fell off. However, they reported no genetic work with this
trait.

Piper and Morse (1923) observed segregation for amount of pubescence in a cross
between a pubescent and a nearly smooth Japanese variety. Johnson and Hollowell (1953)
called this type "appressed hairy" in a report on insect damage study with 27 introductions
of this type, and 7 glabrous ones. William (1950) reported that appressed pubescence

types was found in several introductions from Japan and Korea. He found that the ons
between appressed and normal pubescence types were all intermediate, and that the F3os

gave a 1:2:1 ratio for appressed: intermediate : normal types. Kawahara ( 1963) studied

F2 populations from crosses of a glabrous variety 'Mizukuguri' with pubescent varieties,

'Odate-I' and ’Tansentanryoku' and found that a single dominant gene produced

glabrousness, presumably the P1 of the earlier workers. He also reported a gene pair
'Wewe' for strongly shiny versus weakly shiny leaves. However, since 'We' was 100%
linked to P1, this apparent leaf shine was probably due to the absence of pubescence, and
therefore, explainable by P1 .

Bernard and Singh (1969) studied the inheritance of pubescence in great detail.
They crossed normal type with glabrous (P 1) , curly (Pc), dense (Pd), sparse (P3), and
puberulent (P2), and studied the F2 and F3 data. They found that each of these five

pubescence types differed from the normal by a single gene pair. The normal was dominant

to puberulent, but recessive to glabrous, dense and sparse, and intermediate with curly

pubescence. Four of these types (P1, Pc, Pd, PS) occur in varieties from eastern Asia, and

the fifth (P2) originated as a mutant found in Iowa in 1924. They also studied linkage and

9
allelism among P1, Pc,Pd, and Ps by crossing their isogenic lines in 'Clark' and 'Harosoy'
genetic backgrounds with a puberulent line( T-31). The F2 results showed that the genes
P1, Pc, Pd, and Ps are separate, unlinked or not closely linked loci. The F2 data of the
crosses combining T-3l with the 'Harosoy' isogenic lines, each carrying one of the four
other pubescence genes, also showed that P2 was also a distinct unlinked locus. They
observed that glabrous (P1) appeared to be epistatic to the other types, although Pd and Ps

affected the density of the hair stubs visible on close inspection of the glabrous plants. The
genes pc and p2 and pc affect the form of hairs independently of the density effects of Pd

and Ps. Pd and Ps interact with each other in an additive fashion in controlling hair density.
They crossed T- 145 (glabrous , P1 r) and Clark (normal p 1R ) and studied the F2 ratio to
estimate the linkage of these two loci. The F2 data gave a maximum likelihood estimate of

0.20.+/- 0.46 crossing over between P 1- and R- loci, which does not deviate significantly

from the 0.18 estimate of Nagai and Saito (1923), ie, the two loci are linked.

2.2.2 Pubescence Color

Tawny (brown) and gray pubescence colors are equally frequent among the most
plant introductions and cultivars of soybean in the USA. Woodworth (1921) reported that
pubescence color is controlled by a single gene pair, with tawny (T) being dominant over
gray (t). He also observed that this gene interact with It' (light/dark hilum), W, W]
(purple/white flower), and Rr (black/brown seed coat) genes to give a new hilum color.
Their interaction with the allele '1" produces only gray or black pigments in the seed, while
't' produces imperfect black or buff pigment

10

G I. I . l' E. I I' . I
I_R_T_w1w1 Gray

I__R_tt w] w] Buff

ii R_T_W1__ Black

ii R_tt W1_ Imperfect black
ii R_T_w1 w] Black

ii R_tt w 1w 1 Buff

ii rr T- Brown

ii rr tt Buff

In tawny pubescent genotypes, the trichomes on the young plants are colorless,
but after several weeks of growth, the uichomes on the stems, pods, and leaves develop
brown pigments (Palmer and Kilen, 1987). This pigment is retained in the plant and
facilitates classification of tawny and gray pubescent genotypes. Among gray pubescent
genotypes, most trichomes are without brown pigment, giving a distinct phenotype to the
plants. The '1" allele has a major effect on the production or regulation of an enzyme
necessary for the formation of quercetin from kaempherol. Cultivars with 7" allele have
free quercetin (the aglycone) in the pubescence and those with 't' have free kaempherol
(Buttery and Buzzell, 1973). Bernard (1975) described another major gene pair affecting
pubescence color. The alleles ‘Td' produces dark-tawny, and 'td’ light tawny in presence
of 7' allele. In contrast to "1" allele, which affects the hilum color in seed and flavonol
glycosides, 'T d’ affects only pubescence color. In presence of 'tdtd', there is no or

markedly less flavonol in the pubescence (Buttery and Buzzell, 1973).

11

2.3 Effects of Pubescence on Insects Resistance

Pubescence density in soybean has been found to affect insects resistance. P005
(1929) reported that among 15 species of Homoptera, only Empoascafabae could cause
injury to soybean, and that the extent of injury was related to the amount and kind of hairs
present. Glabrous cultivars were injured much more than the pubescent ones. P003 and
Smith (1931) reported that a glabrous soybean variety showed a greater infestation and
oviposition of leaflloppers (Empoasca fabae) than the pubescent varieties. A number of
introduced soybean varieties were studied at Arlington, Virginia, by Johnson and
Hollowell (1935), where they found severe potato leafhopper infestation and damage to the
glabrous types. The damage were less in cmly types, while the normal pubescent varieties

were undamaged or slightly infested. They also studied the F3, F4, and F5 generations

from a cross between pubescent and glabrous soybeans and observed that glabrous plants
were damaged by leafhopper. Genetic linkage between P1 (glabrous) and genes for
leafhopper susceptibility was ruled out since no cross-over types were detected in a large
population. However, Morse and Carter (1937) reported that Japanese investigators have
found glabrous soybeans to be highly resistant to soybean pod-borer (Laspeyresia
glycim'vorella Mats), while pubescent types are highly susceptible. Both insects are
present in Japan, but only leafhoppers in the USA.

Wolfenbarger and Sleesman ( 1963) obtained seeds of several pubescent types
from the USA soybean laboratory and studied their reaction to leafllopper at two sites in
Ohio. They observed that dense and normal lines had high resistance to leafhopper, sparse
had only low resistance, while the glabrous and curly types showed severe stunting and
hopper burn. Hartwig and Edwards (1970) reported that seed yield of the glabrous type
was significantly lower than the normal types in three of the six years of testing. In those
years, glabrous lines had higher leafllopper damage. They also found that curly lines had
lower leafhopper resistance than the normal ones.

12

Singh et a1. (1971) studied the effect of near isogenic lines of the pubescent types
in 'Harosoy' and ‘Clark' genetic backgrounds to compare leafhopper injury. They found
that the trend was similar in both, but differed considerably with the pubescent types.
Dense pubescent types showed significantly the lowest hopper number followed by normal
and sparse which did not differ significantly. The hopper numbers were highest in curly
and glabrous, but they did not differ significantly among themselves. Broersma et a1.
(1972) developed isogenic lines of pubescent types into 'Clark' and 'Wayne' backgrounds
and studied their effects on leafhopper injury. They found that the orientation of hairs was
more important than the number of hairs for resistance. Glabrous lines had more damage
than others. They believed that orientation and perhaps the size and other hair
characteristics were significant factors in determining leafhopper resistance. Also, they
observed that glabrous strains had significant increase in yield, number of pods/node,
pOdSImain stem, and weight of 100 seeds when the leafhoppers were controlled with an

insecticide.

2.4 Effects of pubescence on Agronomic Traits

Pubescence type and density have been found to affect plant vigor, insect
resistance, agronomic and physiological traits. Hartwig and Edwards (1970) measured the
effects of several morphological traits on seed yield in soybeans by transferring each of the
trait into a common background through backcrossing. The only traits that influenced yield
were indeterminate growth and glabrousness. The lower yield of indeterminate types was
considered due to lodging and that of glabrous types was due to increased damage by the
potato leafhoppers. The seed yield of the glabrous type was significantly lower than the
normal isogenic line particularly in the year when leafhopper damage was severe.
However, yield differences were not significant in the isogenic lines of curly-normal or

dense-normal. The pubescence color had no effect on yield.

l3

Singh et al. (1971) studied the effect of different pubescence types on different
agronomic traits under different environments. They developed near isogenic lines of
normal, dense, sparse, curly, and glabrous pubescence types in 'Harosoy' and 'Clark'
genetic backgrounds. Effects were found to be similar in both the genetic backgrounds.
Dense pubescent plants grew tallest, followed in descending order by normal, sparse,
curly, and glabrous. Yields of lines with normal, dense and sparse pubescence were
similar, and superior to curly and glabrous lines. In 'Clark' background, dense lines
yielded significantly lower than normal , which may be due to lodging of dense lines
during the grain filling stage. The maturity didn't differ among the pubescence types in
'Clark' background, while in 'Harosoy', glabrous was significantly later than other types.
The difference was not significant among normal, dense, sparse, and curly types. The
traits, 100 seed weight, and percent protein and oil in seed were not affected by the
pubescence types . They indicated that the growth differences possibly reflect the action of
genes closely linked with the pubescence genes, or they might result from pleiotropic
effects of the pubescence genes themselves.

Broersma et a1. (1972) studied the effects of pubescence types in 'Harosoy' and
'Clark' backgrounds on leafhopper incidence and yield. They found that glabrous and
curly types had significantly higher hopper number and generally lower yields than other
types. However, when leafhopper incidence was only in the early stage, glabrous and
curly types regained growth , and the yields were identical to the other types. Also, they
studied the effects of leafllopper control with an insecticide, dimethiate in 'Harosoy’
background. They found that yield, number of pods/node, and 100 seed weight increased
significantly in the glabrous lines when insects were controlled. However, there were no
significant difference among curly and normal lines for these traits. Hartung et al. (1980)
studied the effect of various alleles including those for pubescence types in 'Clark' and

'Harosoy' backgrounds. They found that the Pd (dense pubescence) allele had no

14

significant effect on yield averaged over cultivars. However, there was a significant yield
increase in 'Harosoy', but it decreased nonsignificantly in 'Clark' with a decrease in
pubescence density. The Pd allele resulted in more vigorous plants, that were significantly
taller, more prone to lodging, and later in maturity. The increase in yield of 'Harosoy', but
decrease in 'Clark' indicated a significant Pd allele x genotype interaction. Based on their
observation, a synergistic effect of Pd and t (gray pubescence) was present. Sparse (Ps)
had no effect on yield , but significantly reduced plant height and seed weight. The allele
causes a loss in vigor perhaps of the Opposite effects of reduced pubescence as described
for Pd. Semi-sparse (Pss ) allele, however, significantly reduced yields, hastened
maturity, and deceased plant height. The complementary recessive alleles pa] and pa2 (
appressed pubescence) slightly increased yields. However, the influence of the genetic
background was evident, since they significantly increased yields in 'Clark', and decreased
yields in 'Harosoy'. Maturity was also hastened, plant height was decreased slightly,
lodging was increased, and seed quality improved. Again these overall effects arose as a
result of the major effect of these alleles in 'Harosoy' and negligible effects in 'Clark'.
Apparently, the effects of palpaZ gene in the 'Harosoy' genetic background is not clear.
Cooper and Waranyuwat (1985) studied the effects of three genes Pd (dense
pubescence), Rpsl (Phytophthora root rot resistance), and In (narrow leaflet) in near
isogenic lines of 'Harosoy' and 'Clark'. In all these comparisons, the addition of the Pd
gene to the indeterminate isolines resulted in a significant increase in plant height and
lodging, and a significant decrease in yield. In absence of lodging, the height of the
determinate isolines was significantly increased in two of the three comparisons, and the
yield was either increased (Harosoy) or the difference was not significant(Clark). This
differential response of the Pd gene in determinate and indeterminate isolines strongly

supports the hypothesis that failure to obtain a significant yield increase, or in this case,

15

getting a yield decrease by addition of the Pd gene to the indeterminate isolines was mainly
due the increased lodging.

Specht et al. (1985) reported that pubescence morphology could be altered by the
various qualitative genes, and that such alteration might improve adaptation to unique
production environment. They evaluated near isogenic lines of 'Clark' and 'Harosoy',
which possessed genes singly or in combination for pubescence morphology'(pa1, pa2,

Pb, Pc, Pd 1' sz, Ps , and P 1) for their agronomic performance. They found that palpaZ

(appressed pubescence) consistently increased seed yield in 'Clark' genetic background ,

but not in 'Harosoy'. The Pd 1 allele (dense) had little effect on seed yield in 'Harosoy',
but reduced yield in 'Clark'. All Other alleles were either deleterious or neutral in their
effects on yield. Only Pd 1 and sz alleles resulted in greater plant height; while the other
alleles had either no effect or reduced plant height. Thus the Pd 1, sz, pa] and pa2 alleles

may thus offer an adaptive advantage in cultivars for certain production environments.

The morphological change which affects the environment of leaf and may benefit
crop productivity is increased leaf pubescence. Gausman and Cardenas (1973) found that
leaf pubescence on detached soybean leaves decreased the reflectance of near infrared
radiation , but had no effect on the reflectance of photosynthetically active radiation. The
resultant effect of additional leaf pubescence has been reported to be a reduction in
transpiration (Woolley,l964; Ghorashy et al.,1971; Ehleringer and Mooney, 1978) by
reducing the radiation load on leaf. Ghorashy et al. (1971) studied the effect of leaf
pubescence on transpiration, photosynthesis, and seed yield of three near isogenic lines of
soybeans, and found that photosynthetic rates and yields were not affected significantly by
pubescence types (normal, dense, glabrous). The transpiration rates of normal and
glabrous lines were the same, and were significantly higher than the dense type. The
isolines differed in vegetative characteristics, shoot weight, root weight, leaf area and plant

height, which may have influenced transpiration rate. According to Waggoner (1966), leaf

16

hair should reduce diffusion of water more than C02, since the boundary layer resistance

constitutes a greater proportion of total resistance to water vapor diffusion than to C02

diffusion. These findings suggests that water use might be reduced without reducing
photosynthetic rate or yield of soybean by increasing pubescence.

Baldocchi et al. (198 3) studied the effect of leaf pubescence on mass and energy
exchange between soybean canopies and atmosphere. They found that additional
pubescence in an isoline of 'Harosoy' decreased latent heat flux (LE) and increased
sensible heat flux from the crop. The net radiation (Rn), turbulent mixing, and C02
exchange over normal and dense lines were similar. No differences were found in internal
plant water potential or stomatal resistance. They suggested that differential partitioning of
Rn by isolines was due to differential penetration of solar radiation into the canopies- -more
solar radiation penetrated into the Hypersoy dense pubescence (HPD) canopy. The C02
water flux ratio (CWFR) was greater in the I-IPD isoline since additional pubescence
reduced LE. This observation suggests that increasing pubescence density improves water

use efficiency.

2.5 Variance Components

A number of agronomic nits in crop plants are influenced by genes at many loci,
causing a variation in the segregating generation to be continuous or quantitative in nature.
When quantitative factors are involved, linkage may affect the inheritance of a trait.
Knowledge of type and amount of gene action, and degree of linkage influencing the
quantitatively inherited traits is important to plant breeders in selecting a suitable breeding
procedure. Most of the currently used breeding methods for manipulating metric traits have
their origin or rationale in quantitative genetics. The development of a more efficient
breeding procedure for quantitative traits is dependent upon a better understanding of the

nature of gene action underlying the inheritance of the quantitative traits (Burton,1987). In

17

soybean, hereditary variance has been partitioned through experiments using materials
generated by nested (hierarchic) or diallel designs. Relationship among these progenies are
equated with components of variance and covariance among generations. This permits
Least Square estimation of genotypic variance into additive, dominance, epistasis and
linkage effects. The magnitude of genetic variance component is unique to the population
from which the components are obtained. These variance components are influenced by
degree of dominance and allele frequency (Falconer, 1981).

Gates et al. (1960) reported linkage of genes controlling quantitative traits in
soybean. They found that linkage was significant for flowering time, height, and yield, but
not for maturity, period from flowering to maturity, seed weight, percent oil, and lodging.
Linkage in components related in form to additive variances was found in all these
characters, while linkage in components related in form to dominance variance was
demonstrated only for plant height. Repulsion linkages predominated for height and yield,
while coupling linkages predominated for flowering time. Brim and Cockerham (1961)
reported that additive component of variance was significant for all the characters studied.
Dominance effect was too little as was expected for the self-fertilizing species. However,
there was considerable amount of additive x additive epistasis. Cockerham (1963) found a
significant dominance variance for seed size and plant height. However, Hanson et al.
(1967) reported a considerable additive x additive epistasis for seed yield, maturity, and
percent seed coat mottling, accounting for more than 50% of the total genetic variance.

Croissant and Torrie (1971) reported evidence of non-additive effects and linkage
in two hybrid populations of soybeans. They studied F4, F5, F6, and F7 generations in

the first year, and their respective F5, F6, F7, and F8 generations in the second year.

I

Their nested design allowed an estimation of genotypic variance and covariance. Based on
multiple and partial regression analysis, they found that additive genetic variance was the

major component of the genotypic variance for all the character studied. However, they

18

also found dominance variance for plant height, seed weight, and lodging, but there were
relatively small. Linkage components appeared to be important for days to flowering, plant
height, seed weight, and lodging. Brim(1973) mentioned that additive variance is the
primary component of the genotypic variance for the traits of economic importance in
soybeans. Dominance and epistasis may be present with positive and negative signs,
which will tend to cancel out when averaged over loci. However, hybrid vigor and
inbreeding depression have been observed in soybeans, indicating that it may be
worthwhile to look for heterozygous gene combinations. When additive x additive effects
are important, early generation testing may not be an appropriate selection approach, since
an opportunity must be provided for unique gene combinations to come together.
Cockerham (1983) modified the procedures for interpreting the covariances of self-
fertilizing relatives by using several identity by descent measures in addition to the
inbreeding coefficient (F). This has permitted the development of genetic models with
additive and dominance effects that are general for all gene frequencies. VandeLogt et al.

(1984) studied the components of genetic variance and the effects of linkage for the
quantitative traits in barley. They studied F4, F5, F6, and F7 generations from two

crosses in the first year, and their F5, F6, F 7, and F8 generations in the second year.

Least Square analysis was used to calculate additive, dominance and linkage effects.
Additive genetic variance was important for all the u'aits in both the crosses. However,
dominance variance was also present for heading date, kernel brightness, test weight, and
grain yield, but the estimates of dominance variance may have been inflated by linkage
effects. There was coupling phase linkage for grain yield in both the crosses, and for
heading date in one. The results indicated that breeding procedures which keep linked
blocks of favorable genes intact should be utilized in crosses among adapted barley
genotypes. Powell et al. (1985) reported the effect of two major genes, 'denso '(dwarfin g)

and a locus determining 'daylength’ gene on quantitative traits in barley. They developed

19

inbred lines with and without these traits (isolines) and found that the contribution of these
loci to the estimates of additive genetic variance decreased in following rounds of
recombinations(selfing). This demonstrated that in these cases the association between

major genes and quantitative characters was due to linkage disequilibria.

2.6 Heterosis
While additive variance is the primary component of hereditary variance, non-
additive types of genetic effects can contribute significantly to the variation in some traits of

soybean populations (Brim,1973). Though several workers reported heterosis for yield in

soybean , it is quite difficult to produce large scale F1 seed in practice. Brim and
Cockerham (1961) evaluated Fl's from two crosses and estimated means of F2 to F5

generations. The Fl's were significantly greater than the high parent for yield, height, and

total weight in one cross, and for yield only in the other cross. Heterosis above high parent
averaged 20% percent. Inbreeding depression was neither very consistent nor very great
for the advanced generations.

Weber et al. (1970) measured heterotic response in a large number of crosses (3-
24 plants/cross). Seed yield of F1 hybrids averaged 13.4% greater than the high parent of
the cross. More than 75% of the hybrids exceeded the high parent of the respective cross.
Likewise, Hillsman and Carter (1981) found a 12.9% and 6.2% heterosis for yield over the
midparent and high parent respectively. Nelson and Bernard (1984) studied 37 F 1's over
two years and locations in a replicated test and observed a 7.9% and 3.3% heterosis for
yield over the mid parent and high parent respectively. Evidence clearly shows that given
the proper genetic combinations, high parent heterosis occurs. However, it is not yet clear
how much of this is due to dominance, and how much due to dominance x dominance ,

dominance x additive or due to additive x additive epistasis.

20

2.7 Heritability

In soybean breeding, most of the heritability estimates are made by evaluating a set
of lines in one or more environments, and then from analysis of variance, genotypic and
phenotypic variances are estimated and used to calculate the heritability (Johnson et al.,
1955). Two other methods involve single plant evaluation are:

i) Estimation of genotypic variance in a single environment by subtraction of

non-segregating generations(parent or F1) from segregating generations F2 , F3

etc.(Powers,1955).

i) Parent-offspring regression (Falconer,l960): single plant based on means of

progeny.

Yet another type of estimate is the realized heritability, which is a narrow sense estimate
based on the ratio of selection response to selection differential.

Byth and Caldwell (1969) studied the heritability of yield, maturity, lodging, seed
size, protein, oil, phenotypic score, and early lodging in F6 and F7 generations in three
different environments. They concluded that heritability was relatively consistent across
environments for all the traits except for yield. For yield, it was highest under favorable
growth conditions and lowest in poor environments (drought). Brim (1973) presented a
representative sample of heritability estimates from eight populations, and for nine
quantitative traits that are commonly measured in soybean breeding populations.
Heritability was the lowest (0.03-0.58) for seed yield, and relatively higher for other traits.
These estimates were in close agreement as suggested by Johnson and Bemard(1963).
Shannon et al.(l972) estimated heritability for yield, percent protein, and protein yield in
six populations of F3 lines from crosses between high and low protein lines. The
heritabilities for percent protein were higher than those for yield. Protein yield and seed
yield heritabilities were similar. Predicted progress as a percentage of population mean

from selecting the highest 10% of each population (k=1.76) ranged from 3.3 - 4.7% for

21

percent protein, 0.0 - 10.7% for yield, and 0.0 - 10.7% for protein yield. Shorter et al.
(1976) found heritability for percent protein to be 0.70 and 0.86 for protein yield 0.55 and
0.72, and for percent oil 0.84 and 0.83 respectively in two populations of soybeans.

In two recurrent selection experiments, Brim and Burton (1979) calculated realized
heritability estimates for percent protein of 0.29 and 0.34 over six cycles of selections, and
response per cycle of selection was 0.7 and 1.6% respectively of the base population mean.
In a recurrent mass selection experiment Burton and Brim (1981) estimated realized
heritability for percent oil to be 0.21. Openshaw and Hadley (1984) estimated heritability
in two populations in F3 or F4 generations. They also found a similar result, with
heritabilities of percent protein 0.90 to 0.75, percent oil 0.93 to 0.73, and of yield 0.78 to
0.68 respectively in the two populations. In addition, heritability estimates have been done
for a variety of other traits by several workers (Brim, 1973). Most of these estimates are
broad-sense heritabilities (H), which may have some degree of dominance , and additive x
dominance and / or dominance x dominance interaction components in the genetic
component of variance. In self-pollinated crops like soybean, these non-additive
components of variances are not fixable, and thus the heritability estimates based on
genorypic variance are not very predictable.

Kelly and Bliss (1975) estimated heritabilities of percent seed protein and available
methionine in drybean, and found that the broad-sense heritability ranged from 032-071
for percent protein, 043-056 for percent available methionine, and 0.38-0.60 for

available methionine as percent of protein. However, narrow-sense heritability calculated

by the standard unit regression analysis of F3 and F4 family means on F2 and F3 parental
values ranged from 0.63-0.79 , 0.82-0.89, and 0.82- 0.85 in the F3 generation; and from
0.32-0.61, 0.52-0.87, 0.51-0.81 in the F4 generation for the above three traits

respectively. This clearly indicates that the components non-additive genetic variance

decreases in later generations on selfing. Yiran et al.(1990) calculated the components of

22

genetic variance and heritability in the Davis population of gerbera. They found that the
estimates of narrow sense heritability for flowering time was 0.5 while that of broad-sense
was 0.77. Estimates of component of variance indicated that the major genetic effects
controlling flowering time is additive. However, the dominance component accounted for
28% of the total variance, and the environmental component was 23 percent.

Anderson et al.( 1991) reported the heritability and early generation selection

response for resistance to early and late leaf spot in peanut. Selection based on F2 family

means in the F3 generation via defoliation, infection and sporulation was performed for

early and late leaf spot. They calculated broad-sense, narrow-sense, and realized
heritabilities in the two populations for early and late leaf spot disease for lesion number,
infection rating and defoliation. The estimates were significantly different from one to
another type. In most cases realized estimates were higher than narrow sense heritability
obtained via parent-offspring regression, and in most cases were comparable or higher than
broad-sense estimates. This also indicates the presence of non-additive components of

genetic variance causing higher estimates of heritabilities when calculated on broad-sense.

2.8 Correlation Among Traits

Correlated variation of two characters may be due to the similar genetic causes or
due to similar response to environmental influence (Brim, 1973). The two components of
correlated response may be separated statistically. If genetic correlations are high, attempts
to obtain a response in one character by selecting for an associated character may be
worthwhile. This is especially u'ue , when a character of high economic importance has
low heritability compared with the associated character. Soybean breeders have utilized
correlated response to some degree in selection procedure.

Johnson and Bernard (1963) reported genotypic correlations of a few u'aits with

yield. Yield had a correlation coefficient of 0.4 with maturity, followed in decreasing

23

order by plant height (0.3), seed weight (0.2), percent protein seed(0.2), oil (0.1), and
exhibited no correlation lodging & days to flower. Anand and Torrie (1963), and Kwon
and Torrie (1964) obtained genotypic and phenotypic correlations of these traits with yield.
Their results showed that increased plant height, late maturity, and high lodging were
positively correlated with yield both genotypically and phenotypically. On the other hand,
Byth et al. (1969) found that short plant height and resistance to lodging were associated
with yield in crosses involving indeterminate types. They also found that the association of
yield and maturity varied with the environments. However, the correlation coefficients of
protein and oil with yield varied considerably from cross to cross. In general, correlation of
protein and yield was better than oil and yield. The association of several morphological
characters with yield was investigated by Hartwig and Edwards (1970). They u'ansferred
these characters into a common genetic background by backcrossin g. Only two characters,
indeterminate growth and glabrousness were associated with yield. Both affected yield
adversely; the former was due to early season lodging, the latter was due to injury by the
potato leafllopper(Empoascafabae Harris).

Brim (1973) summarized phenotypic and genotypic correlations between yield and
eight commonly measured traits in soybean. It is evident from the differences among the
correlation coefficients from any particular pair of trait, that significance as well as direction
of correlation depend upon the population in which the traits are measured. Simpson and
Wilcox(1983) also studied phenotypic correlations of yield with other u'aits and reported
that it was significant and positive with height, lodging and maturity in all the four
populations they studied.

There has been interest in the study of correlations between yield and yield
components traits, and between yield and physiological traits. Johnson et al. (1955) found
that genetic correlations between yield and pod number were 0.28 and 0.14, and between

yield and seed size were 0.66 and 0.43 respectively in the two populations they studied. In

24

a group of seven cultivars, Pandey and Torrie (1973) found average correlation of 0.5
between yield and pods per unit area, 0.35 between yield and number of seeds/pod, and
0.04 between yield and seed size. On the other hand, Ecochard and Ravelomanantsoa
(1982) found a genetic correlation of 0.95 between pod number and total yield in a
segregating population of spaced plants.

Buzzell and Buttery (1977) found correlations -0.44 and -0.19 between harvest
index and yield in two populations, while Buttery et al.(1981) reported a positive
relationship between photosynthetic rate per unit leaf area (PA) and yield when measured
40-50 days after planting. However, Ford et al. (1983) found no significant correlation
between yield and photosynthesis (rate of C02 uptake per unit leaf area). The negative
relationship between percent seed protein and percent seed oil is well established (Hanson
et al., 1961; Shorter et al.,1976; Brim and Burton, 1979; Burton and Brim, 1981). Burton
et al. (1982) have shown that percent protein in seed, and methionine content of protein are
not correlated. However, Openshaw and Hadley (1981) found that the correlation between
percent oil and percent sugar in the seed was positive and significant, and the correlation
between percent protein and percent sugar was significant and negative. Also, the
correlations between sugar content and yield were non-significant. In the selection
experiment, increase in the oleic acid fraction of soybean oil led to a correlated decrease in

the linoleic and linolenic acid fractions (Burton et al., 1983).

3 MATERIALS AND METHODS

3.1 Development of Experimental Material
Wells H, a soybean cultivar with normal pubescence (++) was crossed to L62-
0801, a near isogenic variant of Harosoy with dense pubescence (Pde). The parentage

and characteristics of the two parents are given below:

 

 

Characteristics Harosoy (Pde) Wells lI(++)

Parentage HSY #6/ T 207(Pd) WLS #3/ ARK(RPSIC)
Maturity group 2 2

Growth habit indeterminate indeterminate

Flower color pink pink

Pod color gray gray

Pubescence dense normal

Pubescence color gray gray

Seed color yellow yellow

Hilum yellow intermediate

 

The cross was made in 1983 and the F1's were grown in1984 to produce the F2
seeds. The F2's were grown in 1985 space planted, and plants with dense pubescence
(Pde, Pd+) were selected. Rows of selfed progenies of F2 selected plants (F23 families)
were grown in 1986. The F3 rows segregating for Pubescence (progeny of Pd+) were
identified, and dense pubescence plants (Pde, Pd+ ) were again selected from those rows
(F3,4). These F3 , families were grown in 1987 and individual plants from the segregating
families(F3,4 ) were threshed separately (PdPD, Pd+, ++ plants). These F3 ,5 plant rows

were grown in 1988 and families were classified as uniform dense pubescence (Pde),
25

26

segregating for pubescence (Pde, Pd+, ++), and uniform normal pubescence(++).
Uniform rows of dense and normal families were bulked separately. Fifteen F2 dense
plants were selected, and within each F2 , two F3 plants were selected, and within each
F3 one normal and one dense pubescence family were finally bulked Thus a total of 60
lines were generated (30 normal and 30 dense) in a hierarchical design ( Figure 1). These
60 families can be classified as 15 F2-derived, times 2 F3 (F2)- derived, times the 2-
pubescence density within each F3 progeny.

3.2 Field Evaluation
3.2.1 Experimental Design and Procedure

The experimental material consisted of 62 entries including 60 lines developed for
pubescence, and the two parents Harosoy and Wells II. These lines were evaluated in
1989 and 1990 at East Lansing (Ingham) and Britton (Lenawee). In the first year, the
pubescent lines were F6 generation while in the second year they were F7 generation. The
experiments were conducted in Randomized Complete Block Design (RCBD) with two
replications. Each plot consisted of two rows of 3.3 m length 51 cm apart. Seed rate was
maintained at 33 per meter of row length. Seeding was done with a soybean planter.
The two sites E. Lansing and Britton differ considerably for soil types. Britton has a
characteristic clay loam soil, while E. Lansing has loam to Capac loam soils. Fertilizers
were applied @ 134 kg [ha NPK (0:0:60) at Britton and @ 179 kg [ha NPK (6:24:24) at
E. Lansing. In order to control weeds herbicides Scancor @ 0.5 kg /ha + Dual @ 6.5 l/ha

were pro-plant incorporated during tillage.

‘11

N'TI

”'11

$11

27

OX.

Wells 11 L62-0801
(++) (Pde)

O

Heterozygote
(Pd+)

1

OOO.

Homozygous Heterozygous Homozygous

(++) (Pd+) l (Pde)

000.

Homozygous Heterozygous Homozygous

(++) (Pd+) l (Pde)

GOO.

Homozygous Heterozygous Homozygous
(++) (Pd+) (Pde)

l l

 

 

 

 

 

Figure 1. Development of family structure.

28

The experiments were planted and harvested on the following dates at each site:

 

 

1989 1990
Site planting harvesting _ planting harvesting
E. Lansing May 24 Oct.18 May 23 Oct. 29
Britton May 12 Oct. 9 May 11 Oct. 16

 

In the 1989 growing season, there was heavy rainfall in the late May and early June.
However, little damage was done due to timely drainage of rain water out the field. The
1990 growing season was normal at both the sites. There were no major pest or disease
problems except for a few Phytophthora affected plants in 1989 at E. Lansing in some of
the entries. Bacterial pustules was common in most of the late maturing lines, but the

incidence was very low without any significant effect on yield.

3.2.2 Data Recorded
Observations were taken for plant height, maturity, lodging, seed, yield, seed size,
and percentage protein and oil in the seed for the individual plot. The description of
measurements are as follows:

Plant height Height of the randomly selected plants in centimeters measured
from the ground level to the tip of the main stem.

Maturity Number of days after August 31 to reach complete maturity of all
he plants,indicated by senescence of leaves and drying of pods
and stem.

Lodging Recorded on a 1-4 scale,

1 = all plants complete erect at 900 angle

29

2 = plants between >600 to <900 angles
3 = plants between >450 to <600 angles

4 = plants at < 450 angles or completely flat

on the ground
Seed yield Weight of cleaned beans adjusted to 14.5% moisture in ton /ha
Seed size Weight in grams of randomly selected 100 seeds adjusted to

14.5% moisture
% Protein Percent of protein in the seed using NMR (Nuclear magnetic
rasonance) technique

% Oil Percent of oil in the seed using NMR technique

3.3 Statistical Analysis

The data recorded over years and locations were used for various types of analysis
to draw inferences about the materials, estimate components of variances, correlation
among traits, regression coefficients , and heritability of the various traits under different

models. Various SAS (1985) procedures were used to derive these information.

3.3.1 Means Comparison

SAS TI'EST procedure was used to estimate mean, standard deviation, range, and
variances for the various components like year, location, density, entry, and their
interactions for the various characters studied. Also, this procedure provided estimates of
'T' to compare these statistics and drawing inferences about the populations. Proc Means
procedure was also used to compute mean, range, variance, CV, of the data by entry, F2,

F3 (F2) , density, location, and their interactions.

30

3.3.2 Analysis of Variance
The analysis of variance was done using PROC ANOVA procedure of SAS, 1985.

Two models were used to estimate the variance components.

3.3.2.1 General Model ANOVA
Here the total variance was partitioned into year (Y), location (L), Y x L, rep (YL),
entry (E), Y x E, L x E, Y x L x E, and error. These estimates were used to calculate

broad-sense heritability and Genetic advance as compared with the nested model.

Statistical Model:
Yijklm = u + Al + Bj + (AB )ij + R(AB)ijk + T I + (AT)il + (BT)jl + (ABT)ijl

+ Et'jklm

Where, u = general mean
Ai = year effect (Y)
B} = location effect (L)
(AB)ij = Y x L effect
R(AB) ijk = rep effect
T1 = treaunent effect (T)
(AT) 1‘! = effect of Y x T
(BT)jl =effectofoT
(ABT)ijl =effectonxLxE

Eijklm = Environmental or error effects

3.3.2.2 Nested Model ANOVA
Here the analysis of variance was performed in the nested design model, where the

total variance was partitioned into the following variance components:

31

Y, L, R (Y xL), F2, F3 (F2), Density (D), F2xD, F3 (F2) x D, L x F2,
LxF3 (F2), LxD, YxF2, YxF3 (F2), YxD, YxF2xD, YxF3 (F2)x
D,YxLxF2, YxLxF3(F2), YxLxD, LxF2xD, LxF3 (F2) xD,
YxLxF2xD, YxLxF3(F2)xD, Error.

Where, Y = year, L = location, D = Pub. density, R = replication

F2 = F2 derived progenies, and F3 (F2) = F3 derived progenies within the F2.

3.3.2.3 Correction for F-test

The 'F' values produced by the SAS in both the models used only the final error
mean square (MS) to estimate F values for each component line. Since each component
line consists of many more items in addition to the error term (Table 1 & 2) in each model,
the F—values produced by the SAS are not correct and need to be corrected. Based on the
items included in each component line, the error term for each line was recalculated by
addition and subtraction of MS of different lines in the Nested model (Table 3 ). Also, a
corrected error degree of freedom (dfe) was calculated for each component line to test the
significance of 'F' values. Approximate dfe was calculated using the following equation:
Example:

MSe=MS F2+MSLD-MS LF2D
Where, MSe is the estimated error MS for a line to calculate the corrected F value.

M8 on the right are the Mean Squares of respective line in the ANOVA

 

 

MSe2
dfe =
MS F22 MS LD 2 MS 1520 2
+ __ + —
dfF2 deD dele)

Where, dfe = estimated error df for testing the significance of corrected F values.

32

 

 

owe coeoabnsm deem .3
3» No 2+ owe 2222-52-3 2 x w x e .w
m: we? +3» Ne 2+ owe 252.3 A x 0 .h
w» me: + w; No u+ omo 24C2-3 > x O .c
wmbwfi + ma No: +w_ Nob + we» we + omo Tm 5390306 .m
chemo + oNo 2-9» Steam .-
Rmb me + mime + one 2-62-5 A x > .m
€93“ +3 me me + mam? + owe I Sveoueooq .m
AND a: + 3 Nb me + we» N9 + omb T.» CCSo> A
m2m e6 85cm

 

<>Oz< 35¢: —fl.—o~—QU EOE“ Amsmvah—wflnvm ENQE Eauovnxm .fi mu—A—Nr—t

33

 

Ewe $8.2 + emu—meet + emu—meet + E» we cut + em...» mo mt +2 came? + owe

2-32-2

0x4 .2

 

Qmomae + 252:: + omo 2-222222 A mm as. x 2 .:
SE35: + mud—No8»? + 8.» No Q: + maamoGE + owe 2.95.: mm x A .22
came 2 .e + on: Nose + 2.2252 + 2. came? + o~o 2-32-82 2 a A E E .o
game mm 22 + emumo «a 3.. +
2.2 amend: +3.2 Noe: + essence + 2.22an + .5; No me .6 memos. + omo 2-32-22 2 a «a .w
ewe $8»: + came 8 E + emumo G 22 + e» me SE: +3.2 amen... J6me; Non...— +
Ewe $92 + ewe—Nome: + emu—meet + 2.» Ne an... + 3:» No 8. +2 «amour + owe 2-3 532352 225 <-
Qmo a 32+ mamoebe + m: Noe—E Rama-«E + owe 2-98 A «2 2mm .2
Ewe $12.:- mume Nu E: +
m.._~ow.;E + macaw»? + a anew-E: + mi mom-=2: + «2» No GB + max-meg: + one 2&2 «m .n
222322. + oao 23: 25.232 .4
532.: + 2223?. + 2:2 No me. + 25%.... + oNo 2-52-2 2 a r .m
.ND 892 + Emcee—e + 2be~oe~ue + 8—month? + gnomes: + m2» No Q: + mince? + owe 2-2 258283 .N
awe mum-=2 + EmomEE + 2.9%3-26 + E $me + m.» Neg? + m2» Ne Q: + Mahmoud? + owe 2-5 85> .—
mZm .2. 85cm 24

 

.<>OZ< Eco—2 2332 2: 52.... 3sz 8.52.5 ESE 283an

.N 033—.

34

 

uNb

2 names: + cub

29» ND mt +2 MENU“: + cub

3522.2 + 2 Mamba.» + one

22.—moat + 232N282 + 2N...» No Q. +2 Mamba: + 222
22.2% Sat + 22:» an M2. +2 name? + cub

Mamba-22. + one

«2» No $2. + mambgz + 0ND

2m: ~28: + 2 Mamba. + umb

823223 5:2 .2
2222-2222225 2 A 22 2m... x 2 x» 2.2
22.32.222.225 2 x «2 x 2 x r .3
2.82.22.22.22 2 125mm x 2 .8
22.32.222.22 2 x E x 2 2m
2-32-225 2 x 2 x w .22
2.22.222 22 E x 2 x > .22
2.22.225 E x 2 x w .2

2-82-8225 22" 25mm x» .:

 

22 2222.. +2: ~22: + 22: No a. +2 2222222 + 922 2-32-2225 2 x E x w .22

Emu 38: +28 @092 +28» mom-2. + 22» no me. + 28.» ab mt +2 mambmt + own 2-32-5 2 x r .2
m: 222:: + £22222: + 922 2.222225 3.2 E x w .2

8 220822. + m2» NON-=22 + S.» No $2.. + QENDSE + own 2-85-3 mm x > .2

222 22 85cm a

 

.2550 .N 235.

35

Table 3. Estimation of the Error MS (MSe) for the individual
component line in the Nested Design Model.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# Source MS term used

1. Year(Y) 13 +3 - 18

2. Location(L) 10 + 3- 18

3. Y x L 18 + 4 - 25

4. Rep(YL) 25

5. F2 10+6-11+13-14-18+19
6. F3(F2) 11 + 14 -19

7. Pub. Density(D) 8 + 15 - 16 + 12- 21 - 20 + 23
8. F2xD 21+16-23+9-22- 17 +24
9 F3(F2) x D 22 +17 - 24

10. Lsz 18+11-19

11. L x F3(F2) 19

12. LxD 20+21-23

13. Y x F2 14

14. Y x F3(F2) 18

15. Y x D 20 + 16 -23

16 YszxD 23+17-24

17. Y x F3(F2) x D 24

18. Y x L x F2 19

19. Y x L x F3032) 25

20. YxLxD 23

21, LxF2xD 22+23-24

22. L x F3(F2) x D 25

23. Y x L x F2 x D 24

24 Y x F3(F2) x D 25

25 Error from Anova

 

 

 

 

36

From the corrected F tests in the two models, the effects of genotype, F2,F3(F2) ,
Density, and their interactions with year and locations on the various quantitative traits

were determined.

3.3.3 Estimation of Genetic Components of Variances

The nested design components of variances was used to estimate genetic
components of variances (additive, dominance, additive x additive) following the procedure
of Cockerham (1983). Because pubescence density is a factor crossed with the random
effects of F2 and F3 -derived families, the effects of Pd-gene on genetic variance of the
quantitative traits were calculated. Cockerham (1983) used covariance of relatives from
self-fertilization and calculated coefficients of the quadratic components for the various
covariances. He developed the following equations to calculate the coefficients for the

genetic terms:

i) Fg (inbreeding coefficient) = 1- (1/2)g
It is the probability that the two alleles of 'g'(individual) are identical by descent.
ii) Otgg (coancestory coefficient) = (1+ Ft)/2
It is the probability that a random allele from g is identical by descent to a random
allele from g'.
iii) ‘Ytgg' (three gene identity measure) = (Fg+ Ft )/2
It is the probability that the two alleles of g and a random allele of g' are
identical by descent.
iv) Atgg' ( two gene pair identity) = Ft + [( Fg - Ft) ( Fg' - Ft )]l (1- Ft)
It is the probability that two alleles of g and two alleles of g' are identical by
descent.
v) Stgg' (four gene probability) = Ft+ [(Fg - Ft) ( Fg' - Ft )] / [2(1- Ft)]

It is the probability that all the four alleles of g and g' are identical by descent.

37

vi) 5tg + g' = [(I- Fg ) (1- Fg' mm 1- F0]
It is the probability that neither of the two alleles of g, nor g' are identical by

descent, and a random one of two possible gene pairs between g and g' has

both pairs identical by descent.

The genotypic variances can be partitioned into the following genetic

terms:

Gijk = m + aij + aik + dijk
Where, Gijk = genotypic value for an individual with jth and kth allele at ith locus.
m = population mean
019 = additive effects of alleles
d's = dominance effects of alleles

Based on these, the covariance among relatives can be equated into the genetic terms

(Cockerham,l983):
Ctgg' = 26 tgg’ s 2 A + 25 tg + g' 021) + 2(‘Y [gg' + g tgg‘ ) D1 + 8 tgg' D2 +(A

tg.g' - Fg Fg') H

Where, Ctgg' = Covariance of g and g'progenies both originating from
t- th generation
02A = Additive variance over loci
02D = Dominance variance over loci
D1 = covariance of aj and dji
D2 = Variance of dij's
H = Inbreeding depression over loci
With two alleles at each locus, H = 02D, and with all gene frequencies being one-half,

D1= 0, D2 = 0.

38

When translated into formulas for the identity measures, the coefficients for the quadratic

components of Ctgg' are :

firms-m Went

02A 1 + Ft
021) (1+ F ) (1mg)
in
D1 Fg + Fg'+ 2Ft
D2 ELLCEgiQLEgL-EQ
2(1-Ft )
H E_L_(_1-_Ef.l_(_l-_Eg'_)
- Ft

Accordingly, Cockerham (1983) developed the coefficients of the quadratic components of
various covariances (Table 4 ).

The test materials were developed in a nested (hierarchical) design with the
common ancestors in F2 and F3 (F2)- derived families. Nested designs are appropriate for
self-fertilizing species. Initial estimates of nested design components of variance or
covariance are linear functions of covariances of relatives. In 1989 the test materials were
F2 and F3 (F2) derived lines evaluated in F6 , and in 1990 they were in F7 generation.
Accordingly, their respective covariances, Ft , Fg and coefficients of genetic variances
were calculated. The coefficients for 0'2 F2 and 02F3 (F2) were calculated by averaging

their respective genetic variances in F6 and F7 generations as follows:

39

Table 4. Coefficients for quadratic components for various covariances

for the F2 and F3 derived lines in F6 and F7 generations.

 

 

 

tgg' t g Ft Fg 02A 02D GZAA
044026) 0 4 o 15/16 1 1/256 1
1440135) 1 4 1/2 15/16 3/2 3/256 9/4
055(F2,7) o 5 o 31/32 1 1/1024 1
1550233) 1 5 1/2 31/32 3/2 3/ 1024 9/4
Average 62F2;
= [02 (F2,6) + 02 (F27) ] / 2
= (02A + 1/256 021) + oZAA in F5) + (02A + 1/1024021)
+ oZAA in F7) /2
= 02A + 5/2048 02D + OZAA
Average 02F3:
= 102 (F3,6) +02 <F3,7)1/2

Average 62F3 (F2):

(3/2 02A + 3/256 02D + 9/4 022m in F6) + (3/2 02A

+ 3/1024 021) + 9/4 02A». in F7) /2

3/2 02A + 15/2048 02D + 9/4 02A».

[ (021:3 - <1ze in F5) + (02F3 - 02F2 inF7 )/ 2

40

(1/2 02A + 1/128 021) + 5/4 oZAA in F6) + ( 1/2
02A + 1/512 62D + 5/4 02M in F7) /2
1/2 02A + 5/1024 021) + 5/4 cZAA

Using these coefficients, a matrix of coefficients for the genetic variances for the various

component lines was developed (Table 5 ).

3.3.4 Least Square Analysis

Least Square Analysis was performed to estimate the components of genetic
variance. PROC REG procedure of SAS (1985) was used for the estimation. The
assumptions are:

i) Errors are normally distributed

ii) Errors are independent of random variables

iii) Errors have a common variance( no heterogeneity of variance)

Regression Model:

7=CB

ycecco

(cc') -1C'y = ( cc') -1cc' [3
B=<ccr107

Wherefi = column vector (25 x 1) of genetic
components of variance
C= matrix (25 x 25) of coefficients
7 = column vector (25 XI) of Mean Square (MS)
values from the AN OVA of nested design
Weighted Least Square Analysis was done to see the variation in the estimates of
genetic components due to models. The principles underlying Weighted Least Square is

that each observation in "y' is divided by the proportionality factor and the rescaled

41

imbue: can"? 52.2223: 2222hv um: <32 um: 5:322: 2:223: sauna"
2: «on? $2.222": 22:2 "e. 5.52"? << 2.222 "32 22:22" mg 5.52 u ax own n:
00.8;

 

22.22 2282 22202 2: 2. 22.: 2.2.2—Eco 25.552 2: .5.— 822222> 2: 22 2:28:25 22 2.2.52: .m 222,—.

42

58%qu

W ND H mmx
Q 30 H VNK

A no H am.»
< 0ND fl MNK

.HW Nb H —mx
<<_N.OHNNx

Shy—None? Enough
a gong.“ fauna?

among 301%
2%qu << @onw:

cub "on x
Q amount"
ups-E

 

5.4.2.5230 .m 035,—.

43

dependent variables will have equal variance. Rescaling gives weight to each observation
proportional to the reciprocal of its standard deviation. The estimates were the same under
weighted and unweighted Least Square analysis. This might be due to the large number
(25) of variables included in the regression model used. Hence, Principal component
Analysis was done to select the important variables from the matrix of coefficients.
PRINCOMP procedure of SAS (1985) was used for this purpose, and the important
variables were selected. Weighted Least Square analysis for each trait studied was then

applied with the selected variables under the following models:

Model 1: Additive components only
Model 2: Additive and additive x additive components only
Model 3: Additive and dominance components only

3.3.5 Estimation of Heritability

The effectiveness of selection for a trait depends on the relative importance of
genetic and non-genetic factors in the expression of phenotypic differences among
genotypes in a population, a concept referred to as heritability. The heritability of a
character has a major impact on the methods chosen for population improvement,
inbreeding and other aspects of selection. Single plant selection may be effective for a trait
with high heritability, and relan'vely less effective for one with low heritability. The extent
to which replicated testing is required over years, locations or environments will greatly
depend on the type and amount of heritability of the traits under consideration. The two
following types of heritability are most often estimated and used by breeders in deciding
breeding procedure and predicting genetic advance:

Broad sense heritability(H) = 02 G/ 02 P

Narrow sense heritability (112) = <12A/o2 P

44

Where, 02 P = 02 G +02 E
02G=62A+02D+621+62E
02 P is the phenotypic variance
02 G is the genotypic variance
0‘2 E is the error variance
02A is the additive variance
621) is the dominance variance
021 is the interaction or epistatic variance

(oZAxA, 02mm, oZDxD, ozAxAxA ........ )

In the present study, heritability estimates were done according to Rasmusson and
Glass (1967). Only broad sense estimates were done using genotypic and phenotypic
variances from the General AN OVA, and the estimates were compared with the heritability
obtained in different models in the Nested AN OVA. Using the components of genetic
variances obtained from Least Square Analysis narrow sense heritability estimates were
done for model-1 ( only additive) and model-2 ( additive, additive x additive), and broad
sense estimates were done for model-3 ( additive, dominance). Also, the estimates were
compared under full and selected variables conditions in models-2 and 3. The heritability
estimates were used to predict gains from selection (Gs). Since the materials were
evaluated over two years and two locations, following formula was used to calculate the

heritability.

Broad sense heritability(H) = 02 G/02 P
o2 G

 

oZE/RLleLYG/LYwZLommZYG/YmZG

45

Where, R,L,Y represent replication, location and year respectively.

Narrow sense heritability ( h2) = 62A/0'2 P or 02A + cZAA / 02 P

Where, 02A will involve all additive and its interaction with the additive variances,
and 02 P will involve all variances included in the respective model and the error variance.
Here too the respective variance was divided with the components involved like in broad
sense estimate to calculate the phenotypic variance.

Also, Parent-offspring regression procedure as proposed by Lush (1940) was used to
compute heritability. The model used was:
Yi = a + in + ei
Where, Yi = performance of offspring on parent
0 = mean performance of parents evaluated
b = linear regression coefficient
Xi = performance of ith parent
ei = experimental error associated with
measurement of Xi
The heritability estimate was adjusted for inbreeding as suggested by Smith and
Kinman(l965) as follows:
Heritability = b/ 2 rxy
Where, rxy is the coefficient of parentage ( 63/64
between F6 and F7 generations)
Gain from selection was calculated according to Sprague and Federer (1951) using the
following formula:

Gain from Selection (Gs)

Gs=KaPH

46

Where, K= Constant based on selection intensity (2.06 at 5% selection)
(IF = Phenotypic standard deviation
H = Heritability of the trait
Gain from Selection as percent of Mean (Gs %)

Gs%=Gs/Meanx100

3.3.6 Estimation of Correlation and Regression Coefficients
The estimates of correlation coefficients among traits are useful for indirect
selection. The character of ultimate importance in a selection program is referred to as the
primary character and those which influence primary character are referred to as the
secondary characters. For example, yield may be considered as a primary character, and
plant height, lodging, maturity etc. as the secondary characters. The potential value of
indirect selection for secondary character that is quantitatively inherited was summarized by
Falconer(l98l):
iy hy
CRx = M ------------
ix in:
Where, CRx = amount of improvement in the primary character
obtained by indirect selection for secondary character
Rx = amount of improvement obtained by direct selection for
primary character

M = genetic correlation between primary character(x) and
secondary character(y)
iy = selection intensity for secondary character

ix = selection intensity for primary character
hy = square-root of narrow sense h2 of secondary character

hr = square-root of narrow sense h2 of primary character

47

The selection for morphological or physiological characters is of no value if the characters'

performance is not correlated with the primary character.

In the present study, PROC CORR procedure of SAS was used to estimate
Pearson's correlation coefficients. The multiple correlation coefficients among seven
characters studied were calculated by overall genotypic means (AVE), year (Y), location
(L), year x location, pubescence density( D), year x D, location x D, and year x location x
D. The effects of year, location, pubescence density and their interactions on correlation

coefficients were compared.

Regression coefficients also reflects the relationship between the dependent
variable and the independent variables. It is the measure of dependence of the dependent
variable on the independent variable. Thus, an unit change in the independent variable will
bring a change in the dependent variable. Based on the nature and degree of regression
coefficients, the dependent variable can be manipulated in a positive or negative direction.
This information can be used as a tool in manipulating of the traits of interest by the
breeders. The model is:

Y=a+b1x1 +b2x2 +b3x3 +

Where, Y = performance of dependent variable

a = point of intercept

bl = regression coefficient of x1 independent variable

b2 = regression coefficient of x2 independent variable

b3 = regression coefficient of x3 independent variable

x1, x2 and x3 are the values of respective variables
SAS (1985) Stepwise REG procedure was used to calculate the regression coefficients for
the dependent variables yield, protein, and oil. Using these coefficients, prediction

equations were developed for the above characters.

4 RESULTS

4.1 Comparison of Treatment Means

The overall mean, range, variance and CV. of the progenies for the seven
characters studied are presented in Table 6. The overall mean for yield was 3.36 t/ha with a
range 0.99-5.03 t/ha, and a CV. of 22.0 percent. Maturity had an overall mean of 30.3
days (after 8/31) with a CV. of 20.7 percent. Plant height exhibited a large variance with a
mean of 103.0‘cm and a CV. of 15.3 percent. Lodging scored on a 1-4 scale had a mean
of 2.2 with the highest CV. of 39.2 percent. Seed size showed a mean of 19.2 g(100
seeds) with arrange of 13.2-26.1 g and a CV. of 11.9 percent. The means for protein and
oil were 39.2 and 19.2 % respectively with very low variance and CV. values. These
statistics for the parents, Wells H and Harosoy also exhibited a similar trend. Wells II had
higher means than Harosoy for yield, while lower for other traits. The CV. were high for
yield, maturity in both, for lodging in Harosoy (Table 6).

The variable means and other statistics when computed by year showed a
considerable variation, and their means were compared using TI'EST (Table 7). The
magnitude and range of difference differed with the characters and year. In general, the
means were higher in 1990 than in 1989. However, variance and CV. showed a mixed
trend. The CV. were higher for lodging (approx. 40%), moderate for seed size, maturity,
plant height and yield (8-24%), and very low for protein and oil (<3.0%). When the
means were compared using TI'EST, the differences were significant for all the characters
except for height. I

The means and other statistics computed by location, showed a considerable effect
of location on the performance of these characters (Table 8). In general, the estimates were
higher at Lenawee than at Ingham . The mean yield at Lenawee was 3.75 t/ha which was
almost a ton higher than at Ingham when averaged over years. Though the difference for

maturity, lodging, and size were not great, TTEST showed a significant difference.
48

49

8~x§o2\§>u§.a>

3% 295% u z

 

 

 

0.8 is 8:... 8.2-”... S... m 33 .5 s
8.8 8.... 8.2 31...... 38 w 3.. 522.. .e
3; z .8 a. .w 98-..... 4.8 w .83 86 Ben .m
8.8 8.2 2.3 93.. .8 w .1883. .e
.8. 8.: 3.8. 383.8 .8. w .ee .85. .m
ed. 2.8: an; 32.8 .8 m 5% has been: 8
a8 3.8 e8... 8.4-8.8 2.. w Angie.» e ..
3

am an an... .82... S: w 33 .5 .5
a.” 8.2 3:. 3.8.8 :8 w 33 522.. .e
2: 8.8 83 88.3.. n: w 623 new Ben .m
S. 8... 8... 3-3 3 . 31:288.. .4
3 :8 8...... o.8-o.8 8.8 w .53 is: .m
3: 8.8 .58 c.8588 «.8 m SS 2.3 been: .8
n: 8.: 5.... more: 8.... w 855...; ..
1.142: u=u~$ .. Hg“

2 $8 :3 .83.: no. 8.. $3 =0 .8
3 8. 8 8.8. 33.8 8.8 8.. $3 522.. .e
e... 3.. 8 08.. .888. we. 8.. 823 new 38 .n
2... 8.? S3 3.-.... m8 8.. :1: 288.. .e
on. 8.8 838 .838 9.8. 8.. as. is: .m
:8 8.8. .88 one... 8 «.8 o? 83 33 base: 8
Q8 8.8 3.3 8.98... 8... 8.. 8.322» ..
flag

.>.o es a> 8E5 8:3. 5e: z .2836 a

 

he «5352.0 .2... Afi .53 :3... .3

6.22226 2: he A.>.DV «2.3.5..
2.3.2. as 3.32...» 62.3....» 69:... .53... =a..o>O .e 925.

50

.8... B 38... 8:88.88 u ..

.... mo 822.5%. n H:

.03qu no 39552 n 2

 

 

 

 

 

 

 

 

.8 W8... .88.... .. 8. 88 .5 8.
58... n8- .8 8.... 8.8.: 8. 98 .5 8.
8.. 88... 8.3-2... .. 8.. 28 532.. 8.
88... 8.8- m8 8:. 8.84.8 8.8 98 532.. 8.
o... 88 Q88... .. 8.8 88 8.8 B8 8.
58... 8.8- e... 8.8 Q88... e... 88 2888 8.
e8 .2... 3.-.... .. 3 98 8.8.5.. 8.
88... 8..- 3... 8:. 3.-.... 8.. 98 8.83 8.
..... e88 o.....-o.8 .8... 28 .83.. 8.
88... 88... me. 28 32-98 8.8. 98 .83.. 8.
8.8 8.8.. c.5188 .. n8 .8 8.2.2 8.
88... 8.8.- e... 8.8 32.8 .8 88 88...). 8.
e8. 8... 88-8.. ... :8 88 85. 8.
88... $8. 3.8 .e... 88-8... 8.. 98 2e.» 8.
EA. E .>.u 85....> 8.5. :82 z 22...... so»

 

53> .3 9.82.3.3 2: he 3:83.38 .23.. 1.... 2:3... .3 5.88:5..0 .5 ~33.

51

.2. mo 0385mm u C.—
83884 n 2mg .8“an u 02—

..83 B 582 88838 u ..
82958 .8 89:52 n Z

 

 

 

 

 

 

 

3 ~38 8.8-8.: 8.2 8“ :0 25

8:8 88.8- v.8 3.8 8.8-8.2 «.2 8m .5 07:
0.8 888.: 8.3-8.8 ”.8 8m 532m 23

£88 83- «.8 8:2 8.21.8 8.8 98 532m 0%
8.2 3.3 8.8+: .. 92 8a 38 88 23

88.8 v8.8- 2: 5% 2.3.2 8: 8m 88 88 GE
EV 88.8 3-3 3 8a 8.883 25

88.8 88.8 «.3. ~88 3-3 .. «a 8m 85883 07:
n2 28 8.83.8 8.8 8m .83 23

88.8 8% wfl $8 8.82-8.8 .. :8 8m 2%: ca
8.8 8.3 8.2-8.8 .. 8.8 8a are“: 25

88.8 83- v: :8 8.8-8.8 8.8 8m 53“: OE
S: 8:. 88-88 .. 8.8 98 22» 2mg

58.8 8.3- 8.8 52 88-8; 88 8m 22» 0E
85 E .>.u 8§E> 8:3. :32 z 288$ 8883

 

A8539—
.3 3822.26 2: .3 8:838 .85: ES 8:3:— uc :cmtanEeU .m via.

$11

We;
01‘)
How
exam

SIIZC‘ F

52

However, the means were not significantly different for protein and oil over the locations.
Likewise, when effects of pubescence density on means and other statistics of the
characters were computed (Table 9), pubescence appeared to affect some of the traits.
However, the degree of difference was not very high. Dense pubescence had higher mean
for maturity, height, lodging and size; almost identical for protein and oil; but lower for
yield than the normal pubescence type. When tested with the 'I'I‘EST, the means were
significantly different for yield, maturity, height, lodging and oil, but non-significant for
size and protein.

The mean and other statistics were further computed on year x location, year x
location x pub. density, pub. density x year , and pub. density x location to see the effects
of these interactions on the performance of the characters. Only the significance of their
mean difference are presented for comparison (Table 10—13). A comparative evaluation of
significance of mean differences due to year, locau'on, and density is presented in Table 10.
It is evident that the means were significantly different for yield, maturity and lodging over
year, location and pubescence density. However, for height the means were not
significantly different over year, but different over location and density. Likewise, seed
size exhibited a significant difference over year and location, but non-significant over
density. Protein means were significantly different only for year, while oil means differed
over year and density.

When the variable means were compared by year x location interaction, the 'l'I'EST
significance were quite interesting. The means were significantly different over locations
both in 1989 and 1990 for yield, height and lodging (Table 11). However, the differences
were non-significant for maturity and seed size in 1989, and for protein in 1990. Mean of
oil percentage did not show significant difference over locations in either of the years.
However, the trend was different when compared over years within the location. For
example, the means were different over years at both locations for maturity, lodging, seed

size, protein and oil, and non-significant over years for yield at Ingham, and for height at

53

.ummunu xe— umfiwflm huddombflwwm I s.

.L. «.0 nouuezum a E.”

60383 Mo 33852 n 2

 

 

 

 

 

 

 

o.a aomo 33.: .. o.a ova :0 38.52

:aoo :a. no nmvo o.aaoé a.o_ ova oo 38o
o.a 8.: o.vv-m.aa Ea ova 532m aéoz

oaooo oa_.o o.a aa.a_ o.vv-v.aa aoa ova 52¢: 3.89
5: 8a a.va-_.v~ a2 ova 0% 88 3882

vfiao mooo a.a~ own ooaaé we ova «2me 3.5a
2v vaoo o.v-o._ o.a ova $33 3882

_ooo.o a? mom goo o.v-o.o .. o.a ova acaofi 3:8
avg aaa o.afloom So ova Eamon 3582

Sooo Ema aé noaa o.aooooo .. nos ova gas: 359
noa Goa oovoga v.oa ova beau: ovéoz

aoooo amaa aoa 33v o.nv.o._a .. a: ova 5532 9.89
o.aa ano mowooo ._. vva ova 22% 3582

mmooo aha- Ea ono 2.33 oaa ova 22» 8:8
EA E .>.u 852$ ease 532 2 «BE; goofing

 

35:28: 3:88:5—
3 282229 2: .3 85:53 .55: .25 252: he gov—«9.8“.» .a ~35.

54

Table 10. Effects of Year, Location and Pubescence Density
on the the significance of variable means using TTEST.

 

 

# Variables Year Location Pub.density
1 . Yield * * *

2. Maturity * * *

3. Height ns * *

4. Lodging * * *

5. Seed size * * ns

6. Protein * ns ns

7. Oil * ns *

 

* = Means significantly different, ns = Means not significantly different.

Table 11. Effects of Year x Location on the significance
of variable means using TTEST.

 

 

# Variables '89 '90 ING LEN
1 . Yield * * ns *

2. Maturity ns * * *

3. Height * * * ns
4. Lodging * * * *

5. Seed size as * "‘ *

6. Protein * ns * *

7. Oil ns ns * *

 

* = Means significantly different. ns= Means not significantly different.

55

Lenawee. When the effects of density on variable means were compared within year and
location, the differences were non—significant in most cases (Table 12). However, the
differences were significant for yield at Ingham in 1989, for maturity at Ingham and
Lenawee in 1990, and for oil at Ingham and Lenawee in 1990. Height was the only
character that exhibited a significant difference over years and location due to pubescence
density.

Similarly, the significance of the effects of density x year and density x location on
the treatment means are presented in Table 13. When the effects of year on density were
compared, the differences were significant over years both in dense and normal types for
maturity, lodging, seed size, protein and oil, but only in dense pubescence affected yield.
Year effects were non-significant for height both in dense and normal types. Similarly,
when the effects of locations were compared in dense and normal types, the variable means
differed significantly in both types over locations for yield, maturity, height and seed size,
but only for lodging in normal type. However, location effects were non-significant both

in dense and normal types for protein and oil.

4.2 Analysis of variance
4.2.1 General Model ANOVA

The AN OVA for General Model for the seven characters studied are presented in
Table 14 & 15 . Here in this model the total variance was partitioned into year, location,
replication, genotype and their interactions. In order to have a more reliable F-test, mean
square for error (MSe) and degree of freedom for error (dfe) for the respective component
line was recalculated (as described in materials and methods) and used to calculate the
corrected 'F' values. With the respective MS and F-significance, C.V., LSD, and grand
mean are also presented in the tables.

The analysis of variance for yield showed a non-significant effect due to year (Y),

location (L) and interaction of genotype (G) with year and location. However, the effects

56

Table 12. Effects of Year x Location x Pub. Density on the
significance of variable means using TTEST.

 

 

 

 

82 9D

# Variable ING LEN LNG LEN
1 . Yield * ns ns ns

2. Maturity ns ns *

3. Height * * *

4. Lodging * * ns *

5. Seed size ns ns ns ns

6. Protein ns ns ns ns

7. Oil ns ns "‘ *

 

* = Means significantly different. ns= Means not significantly different.

Table 13. Effects of Density x Year, and Density x Location
on the significance of variable means using TTEST.

 

 

wt! i x Y W

#Variable Dense Normal m—le
1.Yie1d ns * *
2.Maturity * * * *
3.Height ns ns * *
4.Lodging * * ns *
5.Seed size * * * *
6.Protein * * ns ns
7.0il * * ns ns

 

* = Means significantly different. ns= Means not significantly different.

57

of genotype , Y x L and Y x L x G were significant (Table 14). The CV. was 13.3%
which shows that the error variance was not very high for these measurements.
Likewise, the ANOVA for maturity showed a significant F-test for Genotype (G), Y x L,
Y x G, and Y x L x G. The CV. was 5.8 % which indicated a reliable results produced
by the AN OVA (Table 14). Height exhibited a significant F-test due to Genotype, Y x L,
and Y x G. However, the F-test results were different for lodging. In this case, the effects
of year, location, genotype, L x G, and Y x L x G were all significant. The CV. was
slightly higher (26.5%) for this. The analysis of variance for seed size, protein and oil are
presented in Table 15. Seed size produced a different F-test . The differences were non-
significant for year, location, genotype, Y x G, and L x G; and significant for Y x L and Y
x L x G. However, a significant effect due to year, genotype, Y x L, Y x G, L x G, and Y
x L x G were observed in case of protein . The F-test for oil also indicated a significant

effect of year, genotype, and Y x L x G.

4.2.2 Nested Model ANOVA

In the Nested Model ANOVA, the design partitioned the total variance into 25
components involving year, location, replication (Y L), F2, F3 (F2), pubescence density
(D), and their respective interactions (Table 16-17). Since PROC ANOVA procedure of
SAS (1985) produced the F- values using a common error variance and df, the corrected
error variance (MSe) and degree of freedom for error (dfe) for each component line was
recalculated to compute the corrected F-values and their significance (explained in
Materials & Methods). The analysis of variance showed a differential F-values and their
significance for the seven traits studied (Table 16-17). ‘

The ANOVA for yield, maturity, height and lodging are presented in Table 16.
The F—test for yield showed a significant effect due to Y x F2, Y x F3 (F2) x D, Y x L x

F3 (F2), and Y x L x D, and all other main and interaction effects were non-significant.

Though, density had no main effect, it exhibited interaction effect in some cases. The

58

Table 14. Estimates of Mean Squares (MS) and F-tests
significance of the characters in the General Model

 

 

 

ANOVA.
# Source df Yield Maturity Height Lodging
1. Year(Y) 1 6.03 4931.6 285 .0 44.76*
2. Location(L) 1 82.02 2620.2 8210.3 8.00**
3. Y x L 1 3.72** 2729.3 ** 1128.0“ 0.00
4. Rep(YL) 4 2.71** 28.7 ** 2504.3** 9.71**
5. Genotype(G) 61 0.89“ 101.1** 1137.4" 3.49**
6. Y x G 61 0.40 14.9* 137.6** 0.45
7. L x G 61 0.35 9.4 64.9 0.70*
8. Y x L x G 61 0.32 ** 9.6** 68.0 0.42*
9. Error 244 0.20 3.1 69.5 0.34
Mean 3.35 t/ha 30.2 days 102.8 cm 2.2
C.V(%) 13.3 5 .8 8.1 26.5
LSD(0.05) 0.87 3.4 16.3 1.1
LSD(0.01) 1.15 4.5 21.5 1.5

 

*, ** Significant at p < 0.05 and p< 0.01, respectively.

59

Table 15. Estimates of Mean Squares (MS) and F-tests
significance of the characters in the General Model
ANOVA.

 

 

 

# Source df Seed size Protein oil
1. Year(Y) 1 1379.1 5628.1** 130.20**
2. Location(L) 1 73.4 22.0 0.26
3. Y x L 1 69.8** 8.6"‘* 0.41
4. Rep(YL) 4 22.8** 0.1 0.00
5. Genotype(G) 61 9.6** 3.2** 0.94*
6. Y x G 61 1.0 0.7* 0.51**
7. L x G 61 1.4 0.9* 0.27
8. Y x L x G 61 2.6** 0.5** 0.24**
9. Error 244 0.62 0.006 0.006
Mean 19.2 gm 39.2 % 19.2 %
C.V.(%) 4.1 0.2 0.4
LSD(0.05) 1.5 0.15 0.15
LSD(0.01) 2.0 0.19 0.20

 

*, ** Significant at

p < 0.05 and p< 0.01, respectively.

Table 16. Estimates of Mean Squares (MS) and F-tests
significance of the characters in the Nested Model

60

 

 

 

ANOVA.

# Source df Yield Maturity Height Lodging
1. Year(Y) 1 6.323 4838.7 238.0 48133“
2. Location(L) 1 76.501 2511.6 8151.0 9075*
3. Y x L 1 3.413 2660.2 1267.5 0.000
4. Rep(YL) 4 2.543 26.0** 2410.5 7.762
5. F2 14 2.021 2425* 1904.5* 7.104
6. F3(F2) 15 0.858 74.4* 6624* 3.105*
7. Pub. Density(D) 1 4.193 409.6 18924.2 28.783
8. F2 x D 14 0.255 32.8 338.3 0.954
9. F3(F2) x D 15 0.265 45.4** 491.1“ 1.090
10. L x F2 14 0.609 20.2 81.1 1.313
11. L x F3(F2) 15 0.447 9.6 70.7 0.498
12. L x D 1 0.834 0.0 168.6 0.796
13. Y x F2 14 0878* 30.5 201.0 0.605
14. Y x F3(F2) 15 0.319 13.6 176.3 0.585
15. YxD 1 0.961 121.7 18.0 1.453
16. Y x F2 x D 14 0.174 4.1 98.6 0.274
17. YxF3(F2) xD 15 0.251* 4.8 96.6 0.191
18. Y x L x F2 14 0.544 22.8 49.8 0.587
19. Y x L x F3(F2) 15 0.440“ 11.1** 95.5 0532*
20. Y x L x D 1 1.003* 3.37 453.3" 0.000
21. LxF2xD 14 0.125 4.3 53.3 0.383
22. L x F3(F2) x D 15 0.159 4.5 48.8 0.545
23. YxLxF2xD 14 0.185 3.1 31.6 0.201
24 YxLxF3(F2)xD 15 0.095 3.5 49.0 0.386
25. Error 236 0.205 3.1 68.6 0.328
Mean 3.35 30.3 103.1 2.2
C.V.(%) 13.5 5.8 8.1 26.2

 

*, ** Significant at p< 0.05 and p < 0.01, respectively.

61

estimates for C.V.(13.5%) which indicated that the experimental error with this trait was
very low. The F-test for maturity showed a significant effects due to F2, F3 (F2). and

highly significant effects due to Y x L, F3 (F2) x D, and Y x L x F3 (F2). Pubescence

density did not show much effect on maturity. The estimates of C.V.(5.8%) was within

acceptable range. The F-tests for height, however, showed non-significant effects due to
most of the main and interaction effects (Table 16). The effects were significant for F2, F3
(F2), F3 (F2) x D and Y x L x D. In case of lodging, however, the effects of year,
location, F3 (F2), and Y x L x F3 (F2) were significant. Pubescence density did not
show either the main or interaction effect for lodging. The ANOVA for seed size, protein
and oil are presented in Table 17. Seed size exhibited non-significant F-tests for most of
the components. There were no significant effect of year and location on seed size.
However, the effects of F3 (F2), F3 (F2) x D, Y x L x F2, Y x L x F3 (F2) and Y x L
x F2 x D were significant. Here too, pub. density did not show any direct effect on seed
size. The analysis of variance for protein, however, showed more significant F-tests than
other variables (Table 17). The effects of year (Y), Y x L, F2, L x F3 (F2), Y x F3
(F2) x D, L x F3 (F2) x D and Y x L x F2 x D were significant at 0.05 probability, and
those of Y x L x F2 and Y x L x F3 (F2) x D were significant at 0.01 probability.
Though pubescence density had no main effect in this case, it had considerable interaction
effects. The year effect for oil was significant, and interactions of Y x D, Y x L x F3 (F2),

and Y x L x F3 (F2) x D were also significant. The estimates of CV. in all these cases
were within the acceptable range indicating the reliability of the results.

4.3 Estimation of the Components of Variance
4.3.1 General Model ANOVA
Using the Expected Mean Square (EMS) from Table l and the Mean Square (MS)

from the ANOVA (Tables 14-15) of the respective component line was used to estimate
(:20, cZGY, 026L, 020er , oZe and 02? for the traits studied (Table 18). The

62

Table 17. Estimates of Mean Squares (MS) and F-tests
significance of the characters in the Nested Model

 

 

 

ANOVA.
# Source df Seed size Protein Oil
1 . Year(Y) 1 1327.5 5447.26* 126.58
2. Location(L) 1 70.3 19.76 0.33
3. Y x L 1 63.8 6.96* 0.35
4. Rep(YL) 4 21.1 0.012 0.003
5. F2 14 15.8 9.49* 2.15
6. F3(F2) 15 8.7* 1.80 0.93
7. Pub. Densitym) 1 4.2 3.48 2.72
8. F2 x D 14 7.1 1.47 0.29
9. F3(F2) x D 15 6.2* 0.91 0.38
10. L x F2 14 1.7 1.73 0.55
11. L x F3(F2) 15 1.3 0.89* 0.20
12. L x D 1 6.9 0.00 0.40
13. Y x F2 14 1.4 0.74 1.05
14. Y x F3(F2) 15 0.8 0.76 0.43
15. Y xD I 2.2 0.00 3.88"
16. Y x F2 x D 14 0.7 0.74 0.19
17. Y x F3(F2) x D 15 0.8 0.62* 0.18
18. YxLxF2 14 6.3* 1.12“ 0.36
19. Y x L x F3(F2) 15 2.1“ 0.02“ 0.20“
20. Y x L x D 1 0.42 2.03 0.03
21. L x F2 x D 14 0.95 0.51 0.20
22. L x F3(F2) x D 15 1.33 0.72* 0.20
23. Y x L x F2 x D 14 1.73* 0.68* 0.25
24 YxLxF3(F2)xD 15 0.66 0.21" 0.17“
25. Error 236 0.60 0.006 0.006
Mean 9.2 39.2 19.2
C.V. 4.0 0.2 0.4

 

*, ** Significant at p< 0.05 and p < 0.01, respectively.

63

60:25?» N NO 98 dog u o .omboqonm n m .nowaood u A .30? u .w 695980 N O

 

 

 

 

 

 

 

 

cm ad 59.6 cm ~. ~ aovd vmdoow 0N.N~ >86 numb .0
Boo Boo too oavoo ooo mono maoo eae .m
S “d Ncwd Owad mhmod mud- and N86 AWCND .v
god wood mend- w god mud- Sod- hood AGND .m
wood mvod vowd- mwood a0: 5mm. H cued >060 .N
wmod 8nd fimad ozvd 2.68 mmndfi mood 0ND .H

:0 598m one. “doom £533 Ewen ates: 23> 033; a

 

 

 

 

 

 

 

 

 

.<>OZ< .252

.9555 2: 5?: 92.2.5, be 3:32.58 he 8352mm— .3 «Ban.

 

Tat

64

estimates of 626 was highest for height, moderate for maturity, seed size, lodging and
protein, and relatively low for yield and oil. The trend was nearly similar for the

interaction variance of genotype with year and location. The estimates were negative for
ozGL for maturity, for OZGL and OZGLY for height, and for OZGY and O'2GL for seed

size. The estimates of oze were low in most cases except for height. In computing the

phenotypic variance, negative estimates of components were assumed to be zero.

4.3.2 Nested Model ANOVA

The Least Square analysis of SAS produced the estimates of various components
included in the respective models. Three models, Additive (A), Additive & Additive x
Additive (A & AxA), and Additive & Dominance (A&D) were used to estimate the
components of variance under Weighted and Unweighted Least Square analysis. The SAS
output produced a different estimates for the component lines under additive model for
Weighted and Unweighted analysis. Hence, the Weighted variance estimates were used
and presented here. However, in case of the other two models(A & AxA, A&D) the
computer output did not show any difference in the estimates of variance components under
Weighted and Unweighted analysis. Therefore, Principal Component Analysis (PCA) was
performed for each character for the two models. Based on PCA, important component of
variance were selected for each case to include in Weighted Least Square analysis. When
the selected components were used, the computer results were different for Weighted and
Unweighted analysis. Therefore, for these models the estimates for components of
variance are presented for both full and selected models for comparison and estimates of
heritability. Some of the components of variance produced negative estimates. They were
assumed to be zero while computing the phenotypic variance and heritability.
4.3.2.1 Additive Model '

The estimates of components of variance for the Additive model are presented in

Table 19. The estimates differed considerably with the characters and also with the

65

.8593 H No can :28 u o 62.63 H < 56:2. 00:08.33 n n .5980. u _ .30» u »

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

vaao.o 28o ooooo ooooo eoooo ooooo v_oo.o Seed
oa.v Pa 3.: ad :.a_ aoaa aod no;
ovaamo oovooaa aaoaw oooooo oaoaow- oaoBo oo:o.o aao max 2
oooooo- aovvoo Evoo oammvo- gamma mamooo- aavog _ao aax o
mooooo- maamoo :33 £22 2:2- ovaoEa lama? 1% Ex 2
_oooo.o- aooooo Emaoo mambo baa; oamooo oaoooo ooeao oax E
vooaoo omono anon? ano mavooov maoao avaoo van 5“ 2
833. oavooo 52“an oooooo Eooooa oaoooo- ammoo eao oax a.
oaoooo 58o maovo oosoo manage ovmmfl Svooo soao max :
vaao.o oooo_.o :mvao- aoaaoo boa: aoaeo- aoaooo Sao oax o2
38o aazoo- avvvoo aooaao “Sana- 32a 383W 2% 2x o
vaoo ooomoo Gooao- oa_oo.o aavmmo aoamvar $227 iao EN a
Looaoo mEoo- ooaooo oaaaao moaaav- vaao.o oooooo- Rae 2M a
vaaooo- aooooo aooaoo- asooo stag 388 :83 éaao aca o
Eavoo Eooo osaoo baaoo ooaoou- aoooaa ooovoo saao ox m
vaooo- aaoaoo .voso. omoavaF vmmvoa. ammooo mavaoo 2a ao ox v
mooooo- oaaoo- omomoo- mEoo mama: ammo ovoooo §ao an a
avamoo asno anoaaoo aovaoo- oaana- oavoofi. avzoP siao Q a
Elf have? evmedeemaWLhmmgwa$a .33 m _..

 

 

 

 

 

 

 

 

 

 

 

 

 

.7632 uoEw_o>>A$o>_=cc< 5 353.8.» he 3:28.389 no 8352mm .3 «Ban.

 

66

component lines. The value of 02A were higher than oze for maturity, protein and oil,
and lower for yield, height, lodging and size. The estimates for 02d (Pub. Density) were
generally very low or negative compared to 02A for all the characters except height. The
interaction variances of additive (A) with year (Y), location (L), and pub. density (d) were
lower than their main effects. The estimates of ‘F' and their probabilities are presented for
each for interpretation. The F-tests were also significant for the Weighted Least Square

Regression model used to estimate the components of variance.

4.3.2.2 Additive & Additive x Additive Model

The estimates of variances of the full model are presented in Table 20. The values
were higher for OZAA (Additive x Additive) than 02A (Additive) for all the characters
except protein. The estimates of 62A under this model were either very low or even
negative. When compared with oze (Error) , the estimates of oze were higher than 02M
for all except oil. However, when compared with 02d( Pub.density), the estimates of
ozAA were higher for yield, maturity, lodging, size and oil, and lower for height and
protein. The comparison of year and location variances showed a mixed trend. The
estimates were higher for 021 (Location) than 02y (Year) for yield and height, while lower
for maturity, lodging, size, protein and oil. The interaction variances were lower than their
respective main effects.

The estimates of components of variances under the selected model based on PCA
are presented in Table 21. When PCA was applied to select the important components, the
number and the component line deleted from the full model differed considerably with the
characters. A minimum of two (seed size) and maximum of seven (lodging) components
were deleted in the selected model. The F-test for the Weighted Regression analysis were
all significant. In this model, the estimates of 02A increased compared to the full model,
and were nearly identical for some of the traits like yield, maturity and height. However,

the estimates were negative for lodging, size and oil for 02A, and for protein in the case of

67

.835; H Nb 2... coho u o 63.6.... x 26.2... u << 6.6.2... u < 585.. 8:882... u c £2.82 u _ .82. u a

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

888... 888.8 88% 88.... 88...? 33...... 88...... .% 8.. 8
88.....- 88.... .88... 88.? 88.8 88...? 8.8... .% 8x 8
888.? .88... .88... .88... 888.... E58 ~38... ...% .8. 8
.88.? ~88... 88.... .88... 88.... E8... E8... 3% 8x 8
88.... «8...... .88... 88.... 888.81 88...“ .88... <<% 8.. 8
.88.? 88.... 85.? 88...? .88.: 88: 32...? <% 8.. 8
.88.? 88.... .88.? 8.8...- 88... 88.... 88...... .% 8.. o.
8.8.... 88....- ..88... 88.... 88.8 88....- ..8......- 5.8.. 8.. 8]
.88...- 88.... .88.? 83...”: 8.8.8. 8.8% .88... 6% 8.. t
.88.? 88... 88.... 88...? 88...? 88:? 8.8.? 5% 8.. w.
88...... .82.? E8? .88... 888.9 88.... 88...... 5% 8.. m.
5...... 88...? 88.5 EELS? 888 888.? 3% ENE]
.88... 88.... ~88? .88... .88... .88... 8.8.? 58% ..x m.
88.... 888...- ..88? 88...? 82...? 88.... 888... 1% 8... 8.
.88... 82...? 88.... 8E 88...? ES... 88...? 2% m... :
8.8.... 88.... 88.... 88.... 883.. 88w... 88.... 38% Ex ...
.88.? .88.? 888..- 88.... 82.8.- 8.8.? 88...?! s...% n... a
8.8.8 88.....- 8.8.... 88.... 888.2 88... 9.8.... <5.% 5.
8.8.? 8.8... .88... 28...? 888...- 888.? 88...? s.% a. h
88...? 888. 888? 88...? 88?.- 88.... «8...... 3% EFF.
88.8 888... 88.... 8.8... 838.8 88.... 888... 52% t. m
888.? 8...“...- ....8n? 888? .88... 88...? 8.8.? s..% 8. v
888... 8.8... 88...- 88.... .88....- 88... 8.8.? 5.2% a. m
888.? 88.... 88.... E :88? 88a? 8...... €..% 8. w.
88...... 88...... .88... .. 8...... 888.8 88.... E8... 9% C. .
.5 £22.. 8.88m 2 8.. .86: 8.28,. 22> 28...... 8. ..

 

.982 :5. .22.... 2:83. x 2.83. .2822 g. 85...: .8 25.8.58 8 885...... .8 2......

 

68

89:28.. n no 9... .85 u 9 6.6.2... x 25.2... u << .3228 u < 58:2. 8583.... u c 6288. u _ .89. u »

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

88... 3.8... 88... 88... 88... 88... 8...... 8.32..
Ed VNH No.5 mm.» mad g.n~ 3V.»— 33.n—
888... . 888.8 x 8.8.? 88.... ~88... .% 88 ml
Ila-8.....- 88.... 88.... 88....- E 8.8...- 8.8... .% 8x
.88....- 8888 83.8... 8.88... 88...? 888.8 88.... 8% .8. 8.
x 88...... 88mm... .. SE... . 88.... 88% 8x mm:
~88... 88.....- .88.. 88.... .838. 5.8.. .88... <<% 8.. 8
.88...- 888.7 88...- 28.....- 888.8 8:...“ 5.8... <% Rx ..
.88....- 88.....- 88......- mmmele-RHE «8.8.7 . v% 8... o.
x .38....- 835. :3... x 88.... 88......- <<.% Rx a.
x 8...... L833 88...- . 888..- 28.... 8v% 8.. 5
88.....- ..§..W 88.... 58...- x . ~88... <<.% 8.» or
82...? 83.....- E8...- 58... u x 88.....- <.% 8.. m.
. 8.8... . 5.8....- ...§... . 88...... 2% ex 8.
888.? x 88.....- 8........ 8.8.: 8...... x 58% 8... m.
88.... . 8.8....- E...... 8.8...- 8.8... x <..% 8.. u
3.8.... . x x 88.....- 8.2.... 3.8... 8% m... b
88.... . 3.8.? . .888 88.... 88.... 5.8% Ex .F
88.....- x 28...- x 8.8....- mmmpm..- 88...? 8% a... m
8.8... 88.....- 88...? 88.... 88mm 88.... 8.88... 5.3% :x ..
8.8...- :8... 88m... 88.....- 8.8...- 85..? 88...... s..% a. b
88.....- 88...... Emma... .88.... . .88.... x 83% 8. 8
88.8... 8.8... 388.. . 38.... 888... 888... 5.2% t. m
8.8...- 8.8...- .88...- x 88.... 888.8 8.8...- s..% 8. ..
88.... 8.8... .88.... ..8.....? .8888 8.8... 888..- <<2.% a. m
88.....- 88.... 88.... 8.8.8... 83. 8.8..- «RF... 5.8% «x N
22...“. 8%.... own“... $I§fi 8%.... Li... 8...; WT

 

 

coma.— Eec—z 3.3.8.382 is «3:35. 8 o>=mev< 6.5.25. E

 

 

.20.: $925.4 2.2.2—=80 332:...— .8

3:33.» .3 35:35.5 .3 mouafimfim Au «Sah.

 

69

OZAA. The estimates of oze were identical in both the models, and were higher than 02A
and OZAA in most cases. Here again the estimates of 02d were low, but were positive for
maturity, height and lodging, and negative for size, protein and oil. The variances due to
year and location showed a mixed trend, being positive for some and negative for others.
However, 621 was high for yield and height, and low for size and protein, and negative for
maturity, lodging and oil. Year variances were high for maturity, size and oil, and low for
yield.
4.3.2.3 Additive & Dominance Model

The estimates of components of variances of this model are presented in Table 22.
The estimates of 02D (Dominance) were much higher than 62A (Additive) for all the traits
except protein. Likewise, 62D were too high compared to 0’26 for all the traits studied.
However, the estimates of 62d (Pub. density) were either very low or negative for all the
traits except height. The estimates of 021 (Location) were high for yield, height, and low
for size and protein, and negative for maturity, lodging and oil. The estimates of (fly
(Year) were high for protein, maturity and oil, and low for yield, and negative for height.
The interaction variances were high where dominance (D) was involved. However, this
showed plus and minus effects, and differed considerably with the characters and
component lines. In the selected model (Table 23), the number and type of component
lines deleted from the full model differed considerably with the characters. A minimum of
three and a maximum of eight lines were deleted from the full model. Also, the estimates
of F-test were significant for all the traits. The estimates of components of variances
changed in the selected models compared to the respective full models. However, 026
(Error) remained the same. Here also, the 621) were higher than the estimates of 62A for
all characters. The estimate were negative for 02A in case of seed size and oil. Here
again, the estimates of 02d(Pub. Density) were either very low or negative for most of the
traits. In this case, year variances were not very high for maturity, lodging and protein,

and hence not included in the selected model. However, 02y were high for seed size and

70

.855.» u ab was 52.8 n 9 605588 n O .3225 u < 58:8 85833 n c .2582 n _ 53» n »

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

882 888.8 88% 888.8 E? 33.8.9 «888 »% 8x a
.[Lm88w 888.8 583.8 885 £238 858.? 388 _% 2x F
8888. 888 38.8 5.8.8 282- @855 «3.8.8 1% 2% RI
588.9 88.8 288.8 588.8 882 E88 .888 28% 8x m
#83 $82. $832 85.8 8:3: 833:. 883 9% 8x a
58.9 «88.8 988.8 888.8 38.8 883 888.8 <% Rx 8
88¢ E88 8888. 838 888. 888.8 858.8 8% 8x 8
. 88% 88.2- 885: «88.8 ”8.23: 88%: 83mm- 9% fix 5
828.? 8:3 833 828.9 83.51 28m? 588.8 5% 8x 2
888..- 885.8 3.23,. 88:. 833. 888.8- 383 9% xx E
822? 89.8.? 882W 58.8 88% 338.8 8.88 5% 8x 2
E88 825 208.8 883 883». 28E 38.3 2% ex 5
338 38.8. 883.5 3:.“ 885:“ 285.5. .88.”. 9% :x e
822. 32.8- 3.33- E88. 38%. ~33 58.8 8% Ex a
888.8 2:3- 538.8 3888 83:1 kmtg 888.8. 2% 2x :
2.82 2.8.: 38.8 5:35- 88.88 25.8 888.5 8% 9x 2
088.9 85.9 822- SE; $8.? 58.? 5.8.8. 58% 2x a
Scene 58.2- 880.: R83. 8858“ 888.95 835.5 Q15% 2x bl
588.9 888 83.8 588.8. 358. 38.8 888.9 5% an b
88.? 0888 8888- 85.9 883 85.8.8 822. 3% Br»!
«88.? 888.: 88m: 23.9 88.80 38.8 883 92% ex m
35.? 883- 88%. 88¢ 88: 898- 558.9 3% mx e
088. 5 SE»: 83.8.8- 88“.: 322". 5.2.8 53.5- mg»% «x m
888.9 838? 89.2 258.? 883 88“.? 38.8 5.3% «x m
88.8 888.8 888.8 882 888.8 889: 883 9% 5x _
:0 522a 2: Rom a3 2% been; I22» gang 3“ a.

 

 

.332 ==mAG an 5 8555.5 d 3525. 5 353.3.» he 35.—2.58 he 8353mm .3 939—.

 

71

.8522; H No :5 .85 u 0 60:85:58 u G .2663 u < 58:8: 8583:: n 8 £2.88— " _ .83» u x

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.3ch nix—:5: 28:89:80 .255...—

838 888 888 588 8.88 888 888 £22:
2.8 m8...” 8.» 8.2 88 88.8 ES .8.-
8mum8 x 5288 x 22.8.? x 2288 2% 8x R
8:88- flaw-.2- 9888 822.? 838.8 SR:- 332 _% mm
:888- 52.88 8:2 82.88 8:58- 3538 8888 2% :x R
x 8888. x 2.888 882.8 x 3.888 3% 8x
588 8888 888.188 8228.: £8.53 “BEER.“ :% Lax E
8888- 3888 8888- 8888 88838 838 2.9.8.8 <% Rx
8888. 8888- 8.858- 8888 888.8 8888 a 8% 8x a
8283 SEE 353m 2 538m: 8588- :8% fix .2
8888- 8828 2.52. 288.9 x 85.5 3.88 5% 8x 2
2 85.. 888.2 8588- 85.22 x 5888 9% Ex 2
2888 n 88:88- 8888 $588 x 2.888. 5% 8x B
x 8888- 2 E23 :82- L328 2888- 2% 2x 8
.88: x x x x x x 92% :x 2
L888 2 x x x x x 41% Ex 8
838.: x x 82.88 885.81 2 x 82% 2x 2
85.: 885.: 2 x 58.888 x 888.87 92% ex
8888- 8-588- 2 a 5.8.5. 5.2.8 5538- 52% 2x
2.88.8 888.2 Elm-85.x: 2 8.83.5 8888 92% Ex 8
2888 8288 8:88 8:88- 2 8888 8888 52% x F
2888. x 5888- 3.88- 588 5.88 8288 82% Lax 8
has: 583m 232m x 3388. 8388 x 22% 8x 8
8388- 5:8- 83.28- 2 88B 8888- a 52% x v
IE8: 888-:- ERE: 288.2. 2 32.8: 888.:- 222% x m
2588- x3588 8988 2888 2 new? 3288 5.2% x u
. RES 8888 8888 fimfiFIBbWS 3888 E 0% ELI,
=0 2222 a: warm 2 2 e: 8.522 22% use; SPF!

 

 

:e :32. .252 888.3. A: d 3 8:22:2— d 3225. E 3:32; .8 3:853:88 :8 8.8538”.— .na 938,—.

72

oil. The estimates of 021 were high and positive for height and yield, low for seed size,

while negative for rest of the characters.

4.4 Estimates of Heritability and Gain from Selection
4.4.1 Additive Model

The estimates of heritability (hz, H), Gain from selection(Gs) at 5 % selection
intensity and Gain from selection as percent of mean (Gs %) of the seven characters under
the different models are presented in Tables 24-28. The estimates from the Additive
model are in Table 24 . In this case the heritability estimates are narrow -sense ( hz) which
were computed as the ratio of ozAlozP . The phenotypic variance was calculated as
explained in the chapter materials and methods.

The heritability estimate was highest for maturity (0.86) and lowest for oil (0.36).
For yield h2 was 0.50, and those for protein and seed size were 0.57 each, while for
height and lodging were 0.73 each. The gain from selection showed a similar trend. The
Gs % was highest (33.6 % ) for lodging and lowest for oil (1.1%). Yield showed a
response of 8.0 % , seed size 6.1 % , while maturity and height had 16.4 % and 11.9 %

respectively.

4.4.2 Additive & Additive x Additive Model

The estimates of heritability, Gs and Gs % calculated from this model are
presented in Table 25 . Here also the trend was nearly similar to the Additive model.
However, the estimates were comparatively much lower. Yield showed the lowest
heritability (0.24) followed by protein and oil. Maturity exhibited the highest estimate of h2
(0.49) and those of height, lodging and seed size were around 0.40. The Gs and Gs %
also showed the similar trends. The gain from selection for yield was 0.2 t/ha, and 0.4 %
and 0.2 % for protein and oil respectively. When compared at Gs %, lodging exhibited the
highest gain of 22.2 %, followed by maturity, height, yield, seed size, and protein and oil.

73

Table 24. Estimates of Heritability, Gain from Selection(Gs)
and Gs as percentage of mean(Gs %) in Additive Model.

 

 

# Character h2 Gs Gs %
1. Yield 0.50 0.3 8.0
2. Maturity 0.86 5.0 16.4
3. Height 0.73 12.3 11.9
4. Lodging 0.73 0.7 33.6
5. Seed size 0.57 1.2 6.1

6. Protein 0.57 0.6 1.6
7. Oil 0.36 0.2 1.1

 

h2 = narrow sense heritability.

74

Table 25. Estimates of Heritability, Gain from Selection(Gs)
and Gs as percentage of mean(Gs %) in Additive and
Additive x Additive Model.

 

# Character h2 Gs Gs %

 

Full Model:

1 . Yield 0.24 0.2 5 .4
2. Maturity 0.49 3.1 10.4
3. Height 0.38 7.6 7 .4
4. Lodging 0.39 0.5 22.2
5. Seed size 0.43 1.0 5.0
6. Protein 0.27 0.4 1.1
7. Oil 0.33 0.2 1.1
Selected Model:

1 . Yield 0.27 0.2 5 .0
2. Maturity 0.48 3.1 10.2
3. Height 0.66 10.3 10.0
4. Lodging 0.47 0.5 24.6
5. Seed size 0.30 0.7 . 3.5
6. Protein 0.58 0.8 2.1
7. Oil 0.31 0.2 1.1

 

h2 = narrow sense heritability.

75

When these estimates were calculated from the A & AxA selected model (Table 25), there
were some changes in the estimates. In this case, height exhibited the highest heritability
(0.60) followed by protein (0.58), maturity ( 0.48), lodging ( 0.47), oil (0.31) seed size
(0.30), and yield (0.27). Here the heritability for protein increased considerably compared
to the Additive model (0.27). Yield, however, had the lowest heritability in this case also.

The estimates of Gs and Gs % were influenced accordingly. The Gs for yield was
0.2 t / ha, while for protein and oil were 0.8 % and 0.2% respectively. Height showed a
very high response (10.3 cm). The Gs % showed a response of 24.6 % for lodging,
followed by maturity (10.2 %), height (10.0 %), yield (5.6 %), seed size (3.5 %), protein
(2.1 %) and oil (1.1 %).

4.4.3 Additive and Dominance Model

The estimates of heritability, Gs and Gs % of the characters in full model are
presented in Table 26. Since the estimates of 02A were very low in this model, only
broad sense heritability (H) were computed. Also, the Gs and Gs % were calculated using
the broad sense heritability.

The estimates of heritability were moderate for most of the traits (Table 26).
Lodging exhibited the highest heritability (0.57), followed by seed size (0.45), yield
(0.44), oil (0.37), maturity (0.32), height (0.28) and protein (0.24). The Gs and Gs %
were accordingly influenced. However, these estimates were very high (above 100 %) for
some traits . Also, in the selected model, (Table 26), the trend was almost similar.
However, height had the highest heritability (0.83), followed by lodging (0.63), seed size
(0.49), oil (0.34), protein and maturity (0.33), and yield (0.31). The estimates of Gs

and Gs % were were again very high for lodging, height, seed size, yield and maturity.

76

Table 26. Estimates of Heritability, Gain from Selection(Gs)
and Gs as percentage of mean(Gs %) in Additive
and Dominance Model.

 

 

# Character H Gs Gs %
Full Model:

1. Yield 0.44 3.6 106.1
2. Maturity 0.32 23.7 78.3
3. Height 0.28 66.8 64.8
4. Lodging 0.57 8.5 387.9
5. Seed size 0.45 14.8 77.0
6. Protein 0.24 3.7 9.4
7. Oil 0.37 0.3 17.1
Selected Model:

1. Yield ‘ 0.31 1.9 56.9
2. Maturity 0.33 11.8 38.8
3. Height 0.83 130.8 126.9
4. Lodging 0.63 9.1 412.1
5. Seed size 0.49 14.4 75.1
6. Protein 0.33 4.2 10.7
7. Oil 0.34 3.0 15.4

 

H = broad sense heritability.

77

4.4.4 General Model ANOVA

Also, the estimates of broad-sense heritability (H), Gs and Gs% were computed
using the variances from the General AN OVA Model (Table 27). These estimates were
similar to the Additive model. The heritability was highest for height (0.99), followed by
maturity and lodging (0.88 each), seed size (0.79), protein (0.72), yield (0.62), and oil
(0.46). The estimates of Gs showed a 0.43 t /ha response for yield at 5 % selection
intensity. When compared on the basis of Gs%, it was highest for height (62.7 %),
followed by lodging (56.4 %), while for yield, maturity and seed size it ranged 10-20 %,

and protein and oil exhibited a low response (2-3 %).

4.4.5 Parent-Offspring Regression

Parent-offspring regression analysis was also used to compute heritability and to
compare the results with the other models. Here the regression of '90 means on '89 means
was done by PROC REG of SAS (1985). Further, the adjustments for inbreeding of
parents was done as suggested by Smith and Kinman (1965). The estimates of F-test,
C.V., b, and h2 are presented in Table 28. The regression model was adequate to estimate
reg-coefficient (b). The heritability computed in this way showed a different trend. The
estimates were very low for yield and lodging (0.02 & 0.01 respectively), and moderately
high for height (0.77).

4.5‘ Estimates of Correlation Coefficients
4.5.1 Average Means Correlation

The estimates of Pearson's correlation coefficients (Genotypic) with their
significance are presented in Tables 29-33. The r-coefficients of variables using overall
genotypic means (AVE) are given in Table 29. Yield showed a significant positive
correlation with maturity, seed size and protein, and non-significant with height, lodging

and oil. The estimates, however, was negative with height. Maturity had a significant

78

Table 27. Estimates of Heritability, Gain from Selection(Gs)
and Gs as percentage of mean(Gs %) in General Model

 

 

ANOVA.
#Character H Gs Gs %
1. Yield 0.67 0.43 12.8
2. Maturity 0.88 6.3 20.8
3. Height 0.99 64.7 62.7
4. Lodging 0.88 1.3 56.4
5. Seed size 0.79 1.8 9.2
6. Protein 0.72 1.1 2.7
7. Oil 0.46 0.3 1.8

 

Table 28. Estimates of Heritability by Parent-Offspring
Regression (Regression of 1990 on 1989).

 

 

 

# Character F P>IFI C.V. ' b 112
1. Yield 9974.1 0.0001 21.9 0.037 0.02
2. Maturity 11473.7 0.0001 20.5 0.338 0.17
3. Height 20414.3 0.0001 15.3 1.151 0.77
4. Lodging 2488.8 0.0001 43.9 0.024 0.01
5. Seed size 36006.8 0.0001 11.6 0.215 0.11
6. Protein 69019.8 0.0001 8.3 0.437 0.22
7. Oil 42869l.7 0.0001 3.3 0.215 0.11
h2 = b/2 rxy
= b x 32/63

= 0.51 b

79

Table 29. Estimates of correlation coefficients and their
respective significance among variables using average
genotypic means(AVE).

 

Variable Yield Maturity Height Lodging Seed size Protein Oil

 

Yield _ 0.27** -0.06 0.01 0.38** 0.24** 0.03
Maturity __ 0.28** 0.45** 0.44** O.45** 0.32**
Height _ 0.55“ -0.11* -0.09 ~0.14**
Lodging _ 0.26** O.28** 0.17**
Seed size _ 0.75** 0.50**
Protein __ 0.59**
Oil

 

*,"‘* Significant at p<0.05 and p<0.01, respectively.

80

positive correlation with all other characters. On the contrary, lodging exhibited a very
poor correlation with yield, but significant positive with all other traits. Seed size and
protein also showed a significant positive correlation with all other traits. Oil also had a
positive association with all except yield. These correlation estimates have been used as

AVE in the text and tables.

4.5.2 Effects of Pubescence Density, Year and Location on Correlation

In order to study the effects of pubescence density on the estimates of r-
coefficients, the correlations were generated for pubescence density and the estimates
were compared with the overall or average (AVE) coefficients (Table 30). Here the
estimates in general were in agreement with the overall (AVE) estimates of Table 29. The
association of yield with maturity, seed size, and protein were significant and positive in all
cases; while correlations with lodging and oil were not significant. However, the
association of yield with height were not strong being significant negative in dense, and
significant positive in the normal type. The r-coefficients of maturity with height, lodging,
seed size, protein and oil were significant and positive in all the cases. However, height
showed a strong positive association with lodging, and a weakly negative association with
seed size, protein and oil. The association of lodging with seed size and protein were
significant positive both in dense and normal types, and were in agreement with the AVE
estimates. Moreover, with oil, the correlation was significant positive in normal , and very
weak in the dense type. Likewise, the association. of seed size with protein and oil, and of
protein with oil were all highly significant positive in all the cases.

The correlation coefficients were also computed by year and location, and the
estimates were compared to see how these factors affect the coefficients. The correlation
coefficients by year are presented in (Appendix D.1). Here most of the correlation
coefficients were identical to the overall (AVE) estimates of Table 29 with some

exceptions. The association between yield and maturity was negative in '89, but

81

Table 30. Estimates of correlation coefficients and their respective

significance among variables by Pubescence Density.

 

 

 

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil
Yield

D 0.28" —0.16* 0.04 0.50” 0.31" 0.08

N 0.31” 0.15“ 0.04 0.28" 0.17“ 0.04

AVE __ 0.27” -0.06 0.01 0.38" 0.24" 0.03
Maturity

D 0.29" 0.47" 0.42“ 0.49" 0.35"

N _ 0.20“ 0.39“ 0.46" 0.42“ 0.34"

AVE ____ 0.28" 0.45" 0.44" 0.45“ 0.32"
Height

D 0.58" -0.24** -0.12 -0.12

N 0.42“ -0.01 -0.08 -0.10

AVE _____ 0.55“ -0.11“ -0.09‘ -0.14"‘I
Lodging

D 0.14* 0.23""I 0.08

N 0.38" 0.36" 0.31“

AVE 0.26“ 0.28“ 0.17“
Seed size

D 0.74" 0.46"

N 0.74" 0.56"

AVE __ 0.75“ 0.50“
Protein

D 0.50"

N 0.69"

AVE ____ 0.59"I
Oil

D D: Dense Pub. _

N N: Normal Pub. __

AVE *, *"' Significant at p < 0.05 and p< 0.01, respectively

 

82

significant positive in '90. Likewise, between yield and height was non-significant in '89,
but significant negative in '90. However, with lodging, the association was not significant.
The r-coefficient with seed size and protein were significant and positive, while with oil it
showed significant negative in '89. These coefficients were in full agreement with the AVE
for seed size, protein and lodging, but partial for rest of the characters. The estimates of
maturity with other variables were similar to AVE in most cases and differed for some.
The association of maturity with protein showed a negative correlation in both the years,
however, the AVE had significant positive. Likewise, for oil, the estimates were positive
in one and negative in another year. The correlations of lodging with other variables were
mostly in agreement to the AVE estimates. The r-coefficients were very high between
height and lodging. The association of seed size with other variables were similar in some
and variable in others. Seed size had a strong positive correlation with yield in '90, but
showed a significant negative association in '89. There was almost a negative association
between height and seed size, and also the r-coefficient were non-significant with lodging
and protein in either year. It also showed a negative correlation with oil in ‘90. Protein
on the other hand showed a positive association with yield and seed size in all cases,
however, negative in both years with maturity, height and oil, and no association with
lodging. The association of protein and oil were strongly negative in both the years in
contrast to the overall estimates of Table 29.

The effects of location on the estimates of r-ooefficients are presented in Appendix
D2. The trend was agreeable to the overall estimates (AVE) in some cases, while different
in others. The association of yield and maturity negative at ING and positive at LEN, with
height it was significant positive at LEN and negative at ING. Yield had a positive
association with lodging at LEN. However, the correlation coefficients of yield with seed
size, protein and oil were almost stable and in agreement with the AVE estimates. Seed size
showed a strong positive association with yield, and protein had a significant positive

association at LEN. The association of yield and oil was very low and negative.

83

However, the association of maturity with height were significant and positive in all cases.
Here too, the association of maturity and lodging was very strong and positive. However,
with seed size, protein and oil the r-coefficients were poor and non-significant at ING, and
significant positive in other cases. The correlation of height with lodging were very strong
and positive in all cases, and non-significant with seed size, protein and oil at LEN, and
significant negative in all other cases. Lodging on the other hand, had a positive
association with seed size, protein and oil. Seed size exhibited a strong positive correlation
with protein and oil in all cases. Also, the association of protein and oil was significant and

positive in all cases, contradictory to the year estimates.

4.5.3 Effects of Interactions of Year, Location & Pubescence Density on

Correlation

In addition to the main effects, the interaction effects of Y x L, Y x D, L x D and
Y x L x D on the estimates of r-coefficients were also computed and compared to have a
better understanding of the association among characters, and how they are influenced by
these factors. These estimates were also compared with the overall (AVE) estimates of
Table 29. The effects of Y x D on the estimates of r-coefficients are presented in Table 31.
In this case, the estimates were stable for some and variable for others. For example,
yield showed a strong positive association with seed size and protein, poor or negative with
height, lodging and oil, but variable with maturity. It showed a negative association with
in '89 and positive in '90 both with dense and normal types. The association of maturity
with height and lodging were significant positive and stable in all cases and variable for
seed size, protein and oil. Height also showed a strong positive correlation with lodging,
and those with seed size, protein and oil were weakly negative in general. The association
of lodging with seed size and oil were weak and variable, and nonsignificant (negative)
with protein. The relationship of seed size with protein were mostly significant positive,

and with oil were weak and negative. It was interesting to note that the AVE r— coefficient

84

Table 31. Estimates of correlation coefficients and their respective
significance among variables by Year x Pubescence Density.

 

 

 

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil
Yield
'89 D _ -0.30” -0.15 -0.03 0.51“ 0.57“ -0.29“
N _ -0.18"l 0.35" 0.14 0.26“ 0.39" -0.25"‘
'90 D _ 0.48""I -0.16 -0.02 0.50" 0.15 0.03
N _ 0.59“ -0.15 -0.11 0.37“ 0.27“ -0.14
A __ 0.27” -0.06 0.01 0.38” 0.24" 0.03
Maturity
'89 D _ 0.52" 0.40“ -0.43""I ~0.57“‘I 0.33“
N _ 0.35“ 0.26“ -0.14 -0.33"“ 0.13
'90 D _ 0.35“ 0.43“ 0.21“ -0.16‘ -0.14
N _ 0.26" 0.27” 0.38“ 0.06 -0.08
A ____ 0.28“ 0.45” 0.44” 0.45“ 0.32"
Height
'89 D _ 0.56'“I -0.28" -0.34"‘I 0.06
N _ 0.55"I 0.22“ -0.12 -0.12
'90 D _ 0.68“ -0.32“ -0.22" -0.23"'
N _ 0.48“ -0.18‘ —0.17‘l -0.08
A __ 0.55“ .o.11t .o.09' .o.14"
Lodging
'89 D _ -0.07 -0.12 0.06
N __ 0.21‘ -0.14 0.18“
'90 D _ ~0.09 -0.10 -0.21'I
N _ 0.12 -0.10 —0.10
A __ 0.26“ 0.28“ 0.17“
Seed size
'89 D __ 0.41"" -0.15
N _ 0.13 -0.03
'90 D _ 0.28“ 0.07
N _ 0.34“ -0.07
A __ 0.75” o.so"
Protein
'89 D _ -0.62“
N _ 065""I
'90 D __ -0.S8“
N _ -0.66"
A __ 0.59"
Oil
'89 D _
N D = Dense Pub. _
'90 D N: Normal Pub. __
N *, ” Significant at p < 0.05 and p< 0.01, respectively.

 

85

of seed size and protein with oil were strongly positive. However, when compared by year
and density, the estimates were significant negative with protein , and non-significant
negative with seed size in all the cases.

Likewise, the estimates of L x D also showed an interesting trend(Table 32).
Here the coefficients were in agreement with the overall (AVE) in most cases. However,
association of yield with seed size and protein, maturity with height and lodging, lodging
with seed size, protein and oil, seed size with protein and oil, and protein with oil, were all
fairly stable and positive. It was interesting to note that association of oil with seed size
and protein were very high and positive here in all the cases.

The estimates were further computed by density within location and year (Table
33) and compared with the overall (AVE) estimates of Table 29. Here again, some of the
combinations were stable, while others varied due to interactions of year, location and
pubescence density. As is evident, the association of yield with seed size and protein, and
of maturity with height and lodging were stable in most cases. Also, the r-coefficient
between maturity and protein were almost negative. Similarly, the estimate of height with
lodging were high and significantly positive, and not very stable with seed size, protein
and oil. In this case lodging did not show a strong association with seed size, protein and
oil. The correlation of seed size and protein were fairly high and positive in most cases,
and weak or negative with oil. Here again, the correlation of protein and oil were very
strong and negative in all cases contradictory to the overall (AVE) estimate in Table 29.

The r-coefficients by year x location are presented in Appendix D.3. It is evident
that the estimates were stable for some combinations and variable for others. For example,
the association of yield with height and lodging were very weak and negative in all cases,
and significant positive with seed size in all cases except one. However, the estimates
showed Y x L interaction effects for maturity, protein and oil. The correlation of maturity
with height and lodging were strong and positive in all cases except '89-DIG for lodging.

However, it showed a mixed trend for seed size, protein and oil. It was interesting to note

Table 32. Estimates of Correlation Coefficients and their respective
significance among variables by Location x Pubescence Density.

 

 

 

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil
Yield
ING D _ -0.10 -0.17 0.13 0.55“ 0.31" 0.08
N _ -0.10 0.15 -0.07 0.07 -0.06 —0.17
LEN D _ 0.23""l 0.13 0.12 0.44” 0.41" 0.02
N _ 0.28""I 0.49” 0.34” 0.36“ 0.32" 0.03
AVE ____ 0.27” -0.06 0.01 0.38“ 0.24“I 0.03
Maturity
ING D _ 0.61“ 0.49“ 0.04 0.23" 0.05
N _ 0.59" 0.42“ ~0.03 0.04 -0.02
LEN D _ 0.28“ 0.57" 0.59" 0.74" 0.51”
N __ 0.27” 0.63“ 0.70“ 0.78“ 0.59"
AVE ____ 0.28“ 0.45" 0.44" 0.45" 0.32“
Height
ING D _ 0.45“ -0.31" -0.16 -0.15
N _ 0.24" -0.27"‘I -0.31“ —0.33"'"
LEN D _ 0.68“ -0.12 -0.06 -0.07
N __ 0.55“I 0.22"“‘ 0.13 0.07
AVE __ 0.55" -o.11- .o.o9¢ .o.14”
Wm
ING D _ 0.28""I 0.28“I 0.05
N _ 0.45“ 0.37“ 0.30“
LEN D _ 0.07 0.19‘ 0.12
N _ 0.40" 0.39“ 0.34“
AVE __ 0.26“ 0.28” 0.17“
Seed size
ING D _ 0.73” 0.39“.
N _ 0.67" 0.50"
LEN D _ 0.82" 0.50”
N _ 0.81" 0.60“
AVE __ 0.75” 0.50"
Protein
ING D __ 0.58“
N _ 0.76“
LEN D _ 0.45”
N _ 0.63“
AVE ____ 0.59"
Oil
ING D D = Dense Pub. __
N N: Normal Pub. _
LEN D ING = Ingham _
N LEN = Lenawee _
AVE *, " Significant at p < 0.05 and p< 0.01, respectively.

 

87

Table 33. Estimates of correlation coefficients and their respective
significance among variables by Year x Location x Pubescence

 

 

 

 

Density.
Variable Yield Maturity Height Lodging Seed size Protein Oil
Yield
'89! D _ 047" 030* 000 065** 056** 030**
N __ 029* 026* 0.16 032** 045** 030*
'89L D _ 026* 027* 0.08 037** 037** 040**
N _ 008 073** 029* 029* 023* 023*
'901 D __ 002 0.08 0.18 031* 021 0.04
'901. D _ 0.07 005 000 025* 027* -0.18
N _ 029* 002 024 0.22 036** 040“
AVE _____ o.27** .006 0.01 0.38“ o.24** 0.03
Maturity
'891 D _ 073** 0.00 077** 064** 0.21
N _ 067** 0.19 041** 040** 017
'89L D __ 042** 063** 010 059** 042**
N __ 014 0.34** 0.17 027* 040**
'901 D _ 068** 061** 0.08 017 036**
N _ 065** 0.53** 014 028* 004
'901. D _ 069** 075** 041** 031** 028*
N _ 044** 063** 0.07 0.02 032**
AVE __ o 28** 045" 044” 045” 032”
Height
'891 D _ 0.20 066** 047** 004
N __ 035** 024* 010 034**
'89L D _ 074** 015 011 008
N _ 062** 054** 015 002
'901 D _ 069** 0.07 016 022
N _ 044** 0.01 018 005
'901. D _ 067** 050" 027* 021
N _ 051** 024 013 008
AVE __ 055“ .011 .009 014*-

 

‘, “ Significant at p < 0.05 and p< 0.01, respectively.

88

Table 33. Cont'd.......

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil
Lodging
'891 D _ -0.12 0.12 0.18

N _ 0.16 0.04 0.00
'89L D _ -0.01 -0.23 0.20

N _ 0.22 -0.15 0.32“
'90! D __ 0.40“"l 0.06 -0.20

N __ 0.40” 0.01 -0.03
'90L D __ 041” -0.22 -0.21

N __ 0.00 -0.14 -0.16

AVE __ 0.26“ 0.28” 0.17“
Seed size
'891 D __ 0.55" -0.16

N _ 0.36“ -0.15
'89L D _ 0.20 —0.16

N _ -0.03 0.08
'901 D _ 0.22 0.13

N __ 0.30“ -0.01
'90L D _ 0.37“ 0.08

N _ 0.29‘ —0.17

AVE __ 0.75“ 0.50”
Protein
'891 D __ ~0.65“

N 048“
'89L D __ 072""

N _ -0.87”
'901 D _ .047”

N __ -0.6l ”
'90L D _ -0.63"

N __ -0.73“

AVE __ 0 . 5 9 " ‘
0'1
’891 D I: Ingham __

N L= Lenawee __
'89L D D: Dense Pubescence __

N N: normal Pubescence _
'90] D _

N _
'90L D _

N _

AVE

 

‘, ” Significant at p < 0.05 and p< 0.01, respectively.

89

that the r-coefficients were negative between maturity and protein when computed by
location within year, and were opposite to the overall (AVE) estimates. Further, the
association between height and lodging were all significantly positive, while with protein
and oil were almost always negative, and with seed size it was variable. The association
of lodging with seed size , protein and oil were variable, and those of seed size and protein
significant positive except for '89—LEN. The estimates of seed size with oil were all
weakly negative and non-significant as opposed to the AVE (0.49“). Here again, the
association of protein and oil by year and location were highly significant and negative,

while the AVE were highly significant positive.

4.6 Estimates of Regression Coefficients and Equations

SAS (1985) General and Stepwise Regression procedures were used to estimate
the regression coefficients for the independent variables. In this case, yield, protein and oil
were taken as the dependent variables, and for each the remaining six were considered
independent. The general regression model produced the coefficients for the independent
variables, and an estimate for the intercept. Further, in the stepwise regression model, less
important variables were dropped, and estimates of remaining variables and the intercept
were readjusted. The values of the stepwise procedure were used to develop the predican
equation for the three dependent variables.

The estimates of regression coefficients (b) with their T-test or F-test and
respective probability for the general and stepwise regression for yield, protein and oil are
presented in Tables 34-36. In case of yield, the regression coefficients for height and
protein were not significant in the general model (Table 34). The coefficients were
positive for seed size and maturity, and negative for others. In the stepwise regression,
height and protein were dropped from the model, and estimates of intercept also decreased
slightly. Using the coefficients of stepwise regression, the prediction equation for yield

was developed as follows:

Table 34. Estimates of Regression coefficients by General
and Stepwise Regression Analysis for Yield.

90

 

 

 

 

 

 

 

 

General Stepwise
# Variable b ITI P>ITI b IFI P>IFI
1. Intercept 4.9435 5.14 0.0001 4.7727
2. Maturity 0.0278 4.65 0.0001 0.0268 21.58 0.0001
3. Height -0.0012 -0.48 0.6285 -----------------------
4. Lodging -0.1169 -2.86 0.0044 -0.1289 14.21 0.0002
5. Seed size 0.1468 7.22 0.0001 0.1424 79.24 0.0001
6. Protein -0.0076 -0.53 0.5968 -----------------------
7. Oil -0.2377 -4.50 0.0001 -0.2434 25.17 0.0001
Table 35. Estimates of Regression coefficients by General
and Stepwise Regression Analysis for Protein.

General Stepwise 4
# Variable b ITI P>ITI b IFI P>IFI
1. Intercept -1.2864 -0.42 0.6720 -4.9338
2. Yield -0.0772 -0.52 0.5968 ------------------
3. Maturity 0.0665 3.46 0.0006 0.0665 15.20 0.0001
4. Height 00186 -237 0.0179 -------------------
5. Lodging 0.3262 2.51 0.0125 -------------------
6. Seed size 0.8237 14.59 0.0001 0.8501 623.32 0.0001
7. Oil 1.2485 7.75 0.0001 1.3365 81.89 0.0001

 

b = Regression coefficient.
ITI = Estimates of 'T‘.

P = Probability level.
IFI = Estimates of ".F

91

Yield = 4.7727 + 0.0268 Maturity - 0.1289 Lodging
+ 0.1424 Seed size - 0.2343 Oil

Similarly, the estimates of regression coefficients for protein for the general and
stepwise regression models are presented in Table 35. The estimates of coefficients were
high and positive in the general model for oil, seed size and lodging, low for maturity and
negative for yield and height. However, when stepwise procedure was applied, the
estimate for intercept increased considerably (negative), and only maturity, seed size and oil
were finally included in the model. Based on these, the prediction equation for protein was
developed as follows:

Protein = - 4.9338 + 0.0665 Maturity + 0.8501 Seed size

+ 1.3365 Oil

The estimates for oil under the two models are presented in Table 36. In the
general model, the estimate of intercept was high, and those for maturity, lodging, seed
size and protein were low and positive. When stepwise procedure was applied, lodging
variable was deleted from the model, and the estimates for others were slightly readjusted.
The coefficients for yield was high and negative , and those for maturity, seed size and
protein were low and positive. Based on these, the prediction equation for oil was

developed as follows:

Oil = 15.4194 - 0.1756 Yield + 0.0147 Maturity - 0.0057 Height
+ 0.0517 Seed size + 0.0907 Protein

92

Table 36. Estimates of Regression coefficients by General

and Stepwise Regression Analysis for Oil.

 

 

 

 

General Stepwise .

# Variable b ITI P>ITI b IFI P>IFI
1. Intercept 15.4975 38.87 0.0001 15.4194

2. Yield -0.1730 -4.50 0.0001 -0.1756 9.94 0.0017
3. Maturity 0.0141 2.73 0.0065 0.0147 8.54 0.0038
4. Height -0.0063 -3.02 0.0026 -0.0057 5.32 0.0215
5. Lodging 0.0175 0.49 0.6190 --------------------
6. Size 0.0506 2.79 0.0054 0.0517 1 1.55 0.0007
7. Protein 0.0902 7 .75 0.0001 0.0909 261.10 0.0001

 

b = Regression coefficient.
ITI = Estimates of 'T'.
IFI = Estimates of 'F'.
P = Probability level.

5 DISCUSSION

5.1 Means, Simple Statistics and their Comparison

It is evident from the study of means and other simple statistics that overall
statistics are greatly influenced by the nature of characters, population studied, and the
environmental factors operating. The range, variance as percent of mean and C.V. were
higher for yield, maturity, height and lodging, while lower for seed size, protein and oil.
This may be due to the fact that the two parents used in crossing and developing these
populations differed comparatively more for those traits , and less for seed size, protein and
oil. This observation also reflects the number of genes involved in the control of these
quantitative traits. The greater the number of genes involved, the greater would be the
variance in the population, and also, the environmental variance will be large, resulting in
an increased C.V. in the population. From the current study, it is evident that for the traits
like yield, maturity, height and lodging, there may be larger number of genes involved
compared to seed size, protein and oil. This would indicate a comparative ease or difficulty
in handling these traits through breeding. Based on these current observations , it appears
that it would be easier to manipulate traits like seed size, protein and oil through breeding.
However, in the current population, the genotypic variance is quite low for these traits,
hence response to selection will be low. The more variability in a population, the greater
would be chance of improvement through selection. Therefore, a population with high
mean and high variance would be better for manipulation by breeder than one with high
mean and low variance.

The trend of range, variance and C.V. for these traits were identical in both the
parents and the progeny. This clearly indicates a strong environmental influence on these
quantitative traits. Therefore, it would be essential to evaluate the materials over years and
locations across the environments before drawing conclusions about them. Predictions

based on one year or one location data may not be very reliable. When the ranges and
93

94

variances of the progeny were compared with the parents, they were considerably higher in
the progeny for all the characters, even beyond the parental limits. This may result from
transgressive segregation. This clearly shows one avenue for breeder to exploit and select
for superior types even if such genetic combination is not present in the parental
population. The opportunity of exploiting transgressive segregation will be higher for
quantitative traits where large number of genes are involved. It is evident from Table 6, that
the range for yield is the highest among all the traits including height and maturity. This
also reflects that yield has greater number of genes than any of these traits. The number
of genes and the type of gene action in the control of a trait will determine the breeding
procedure to be used for improvement of that trait.

The computations of these statistics by year, location, pubescence density, and
their interactions also produced the identical trends. A higher range, variance and C.V. for
yield, maturity, height and lodging, and comparatively lower for seed size, protein and oil
supported the overall (AVE) results. A significant TI'EST for all the characters except
height indicated that the means were different due to year. The means were higher in 1990
indicating that the growing season was favorable. During 1989, there was a heavy rainfall
resulting in water logging for a period in late May and early June, which may have
caused the variable results in 1989. However, a non-significant difference in height may
have resulted from genetic factor confounded with other factors like location, density,
environment and their interactions. The year effects were more pronounced for protein, oil
and seed size, and less for other traits, indicating that different quantitative traits respond
differently to the year factor. Therefore, for such traits testing over years is necessary for
making valid inferences about them.

Likewise, a differential response of traits to location indicated a similar finding.
All treatment means except protein and oil differed significantly across locations. This
may be due to the fact that the two locations inter-reacted identically towards these traits, or

the parents did not vary much for these traits. On the other hand, this could also result as

95

the confounding effects of year, pubescence density and other environmental interactions
with location, finally neutralizing the difference. Similarly, the effects of pubescence
density on seed size and protein showed a non- significant difference. This may be
possible that these characters are not influenced by pubescence density, or their effects are
not very great. On the other hand, there may be effects of other factors like year, location
and their interactions, finally neutralizing the effect of pubescence density. However, a
significant effect of pubescence density on yield, matmity, height, lodging and oil with a
differential response indicated that the genes for density have a negative pleiotropic effect
on these traits.

The study of interaction effects of year, location and pubescence density on the
means and other statistics revealed an interesting information. Some characters showed
interaction effects , while others did not. For example, there was no effect of location on
mean oil both in 1989 and 1990. This indicated that there is little need of location testing
for oil. On the other hand, a significant difference in means for height and lodging both in
1989 and 1990 due to location revealed that location testing for these traits is essential.
Likewise, a significant difference over years within a location would indicate a need of
more testing over years to have reliable conclusions.

The effects of pubescence density within year and location exhibited interesting
results. The means did not differ due to pubescence density for seed size and protein at any
location or year. Therefore, for studying the effects of density on these traits, no intensive
evaluation over years and location would be necessary. However, height and lodging
showed almost a significant difference in means due to pubescence density within location
and year. This would suggest that an intensive testing of materials over years and
locations would be essential to draw inferences about the effects of pubescence density on
these characters. Comparing the density types, effects of year and location on the means of
characters were quite interesting. For example, year effects were non-significant for height

in both dense and normal types, while significant for rest of the traits. Therefore, for

96

height, not much testing over years will be required, while for others, testing over years
will be essential. Likewise, a non-significant effect of location for protein and oil in both
the density types, would indicate that location testing will not be essential for these traits in
either pubescence density type. However, effect of location for lodging was not significant
in the dense type, and significant in the normal type. This differential response would
indicate that location testing for lodging though not essential for dense type, will be

important for the normal type in order to make valid conclusions.

5.2 Analysis of Variance

The analysis of variance in the General model ANOVA showed in general a
significant F-test for all the characters due to genotypic effects. This indicates that there is
considerable genotypic differences in the population, and selection for those traits would be
effective. Other main and interaction effects were different for the different characters. For
example, yield and maturity showed a significant effect of Y x L x G indicating that the
performance of the genotypes differed with year and location. Therefore, superior lines
need to be selected separately at each site, and be evaluated over years to find a stable one.
Likewise, a significant Y x G effect for height indicated a more testing over years than on
location for this character. Lodging showed a significant effect due to year, location and
interaction of L x G and Y x L x G. This would indicate that this trait is highly influenced
by year and location. The interaction effects would also indicate that the response of
genotypes to lodging is highly variable over environments. Therefore, location specific
selection is important, and also be based on testing over years within the location. Seed
size, protein and oil, all showed the effects of Y x L x G interaction indicating that selection
for these traits be done on location wise, and be based on several years of testing.

The Nested- Design ANOVA was designed to study the effects of pubescence
density (D) on the quantitative traits and their components of variance. The total variance

was partitioned into 25 components as shown in the model. The responses were different

97

for the different traits. The main effects of pubescence density, year, and location were not
significant for yield. This observation supports the findings of Hartwig and Edwards
(1970), Singh et al.(1971), and Hartung et al. (1980), where they did not find any yield
difference between dense and normal pubescence types. Also, Powell et al.(1985) while
studying the effects of two major genes, denso and daylength on quantitative traits in barley
reported that effects of these genes decreased in the advanced generations. Therefore, this
demonstrated that the association between major genes and quantitative characters was due
to linkage disequilibria. Similar conditions may be operating for pubescence type in
soybean. However, it may be also be possible that the difference in pubescence density
between the two normal and dense types was not adequate to bring about the difference in
the quantitative traits studied. Other environmental factors like drought and insect incidence
were not adequate to bring about a major effect in the present study. Previous workers have
reported the positive response of pubescence density on yield in the determinate varieties of
soybean.

In the present study, the effects of pubescence density with some interactions,
however, were quite pronounced. For example, effects of Y x L x D, Y x F2, Y x F3 (F2)
x D, and Y x L x F3 (F2) were significant for yield which indicated that density affected
yield in a particular year and location only, and not on overall basis. Likewise, pubescence
density showed a significant response in a particular year and F3 (F2) line which suggested
that in the locations or genotypes where pubescence density shows a positive response to
yield, density may be used as an important trait in breeding. In the area or population
where there is no response of pubescence density, it may not be an important trait.
However, several workers ( Wolfenbarger and Sleesman, 1963; Hartwig and Edwards,
1970; Singh et al., 1971; and Broersma et al., 1972) reported a positive association of
pubescence density with the resistance to potato leaf-hopper. Therefore, in the situation,
where insects are the major problems, pubescence density may play an important role

indirectly affecting the quantitative traits. Since no deleterious effects of density gene (Pd)

98

has so far been reported, it would be useful to include this trait in breeding as it provides
resistance to drought and leafliopper.

A significant F2, F3 (F2), F3 (F2) x D, Y x L x F3 (F2), and Y x L effects on
maturity indicated that maturity differed with the genotypes and also due to interaction of
genotypes with year, location and pubescence density. A significant F3 (F2) x D revealed
that density affect maturity in certain F3 (F2), and not in others. A differential response of
density with genotypes has also been reported by Singh et al.(1971). They found no
significant difference in maturity due to pubescence types in Clark isogenic lines, but
significant in Harosoy. Hartung et al. (1980) also found that dense pubescence (Pd) allele
resulted in more Vigorous plants with increased height, lodging, and maturity in Harosoy.
Most of these results are however, based on the study of a few genotypes. Therefore, it
seems important to study the effects of pubescence density over a range of diverse
genotypes in order to make valid conclusions.

Similarly, significant effects of F2, F3 (F2), F3 (F2) x D, and Y x L x D on plant
height clearly reflect the importance of density. Plant height is influenced by genotype and
its interactions with pubescence density. Also, a significant interaction of year and location
with pubescence density reveal that the response of density will vary with year and location
for height. Therefore, it is important to identify the location where density has response,
and test over years at that location to make reliable conclusions. There was no main or
interaction effects of pubescence density on lodging in this study. This finding is in
agreement with Singh et al.(1971) and Hartung et al (1980) where they did not report any
significant difference in lodging between dense and normal types. However, the dense
types had increased height and lodging compared to normal types. Hence, it is evident that
pubescence density may not be important or an essential trait in breeding for lodging
resistance. However, a significant year and location effects indicate that the genotype will

show a differential response to lodging over years and locations. Therefore, an extensive

99

evaluation over years and locations is essential to draw a valid conclusion about the
genotypes to lodging.

The ANOVA for seed size showed a significant F-test due to F3 (F2), F3 (F2) x D
and some higher order interactions (Table 17). A non- significant F-test of F2 indicates that
the F2-derived lines did not differ for seed size , however, the F3 (F2)-derived lines
differed significantly. Hence, selection in F3 (F2) derived families would be effective.
Also, a significant F3 (F2) x D revealed that pubescence density affected seed size in
certain genotypes , and not in all. This result supports the findings of Singh et al.(1971)
and Hartung et al. (1980). They also reported an increase in seed weight with pubescence
density. Therefore, it would be essential to identify genotypes which show response to
density for seed size, and use them in breeding or selection for increased seed size. A
higher order interaction of F2, F3 (F2) with Year and location revealed that the genotypes
differed in their response to seed size over years and locations. Also, an interaction Y x L
x F2 x D revealed that there was differential effects of pubescence density over years,
locations and F2's. Therefore, it would be important to evaluate F2's-derived lines over
years and locations to detect the effects pubescence density on seed size.

The F-test for protein showed a significant effect due to year, Y x L, and F2
indicating that the variation in performance of genotypes was great, and needed a testing of
materials over years to draw a valid conclusion. There was no direct effect of pubescence
density on protein. However, significant higher order interaction of Y x L, F2, F3 (F2),
and density revealed that there were differential response of density on genotypes, year and
location. Therefore, it is important to identify a genotype that respond to pubescence
density for protein. Such genotypes should be evaluated over years and locations to finally
identify a superior genotype with a stable response of density to protein. Singh et
al.(1971) also observed an effect of pubescence density on protein in Harosoy, being non-
significant among dense, normal, sparse and curly, and significantly higher in the

glabrous. These findings suggest that pubescence density does not show a positive and

100

strong effect on protein. The comparison of density means (Table 9) also did not show any
significant difference for protein. This may be possible that the genes for density do not
affect significantly the path way of protein synthesis, and indirect selection for protein
through pubescence density will not be effective.

The analysis of variance for oil showed a significant effect due to year indicating
that genotypes must be evaluated over years to make a reliable interpretation. There were
no main effects of F2, F3 (F2), or density in this case. This indicated that there were no
genotypic variation for oil, which may be possible if the two parents used were identical for
oil content, and there were no transgressive segregation for this trait in the progenies to
make a significant difference. Also, there may be very high environmental influence in the
expression of this trait, and the real genotypic difference being very low. Therefore,
selection for high oil will not be effective in this case. A significant Y x D effect revealed
that pubescence density affect oil in some years, and not in others. In such situation,
materials should be evaluated over years to make any valid conclusions. A significant Y x L
x F3 (F2) x D effect indicate that density affect oil in certain genotype, location and year
only. Selection for high oil should be done in location and year where the effect is
significant. In general, there is no significant effect of pubescence density on oil. Singh et
al.(1971) also could not find any significant difference of five pubescence types (normal,
dense, sparse, curly, and glabrous) on oil percent in Harosoy isolines. However, a more
detailed study of the effects of pubescence density on oil on a wide range of germplasm is
necessary to draw a vital inference. When the effects of density were separated by year and
location(T able 12), the effect of density on oil were non-significant in 1989, but significant
in 1990 at both locations. This clearly reflects a greater environmental influence in the
expression of the oil percentage than the density factor alone. Therefore, selection should

be based on the average performance of lines over years and locations.

101

5.3 Estimation of Components of Variances

5.3.1 General Model ANOVA
The general ANOVA model provided the estimates of 02G (Genotypic) and

interaction of variance 02G with year (Y) and location (L). Also, the estimates of oze
(Error) was obtained. The estimates of 62G were higher than cze for all the characters
which provided the evidence that 02G was the major component of variance for the total
variance. The lower estimates of O’ZY and 02L in most cases indicated that these effects
were comparatively much lower than the genotypic effects. As clear from the Table 18, the
genotypic variance for yield and oil were much lower which indicated that selection for
these traits will not be very effective in this population. A comparatively much higher
variance for maturity, height, lodging and seed size would indicate a better response to
selection for these traits in this population. However, this model did not provide
information on the effects of pubescence density on the quantitative traits. This was a good
model to estimate the genotypic effects and its interactions with year and location, and to

estimate broad sense heritability.

5.3.2 Nested Model ANOVA

The estimates of variance components in the Nested Design ANOVA provided
variable estimates under the three different models: Additive(Model—l), Additive & Additive
x Additive(Model-Z), Additive & Dominance(Model-3). Also, the estimates differed
slightly when selected variables based on Principal Component Analysis(PCA) were used
in the selected model.

5.3.2.1 Additive Model
In this model the total variance was partitioned into 17 components , which did not
include additive x additive (AA) and dominance (D) components of variance. A

considerable variation in the esdmates of components among seven characters indicated that

102

the effect of each component was different for different traits. For example, estimates of
oze were much larger for height and maturity, and lower for protein and oil. This
indicates that height and maturity are under control of comparatively greater number of
genes than protein and oil. This could also arise if errors in measurement of height and
maturity were more than protein and oil. In general, as the number of genes increase, the
genotypic, phenotypic and error variances increase. Hence, these variations in the
population or among traits, may dictate the breeding procedure. If the number of genes
involved are just a few( eg. kernel color in wheat), breeding and selection would be much
easier than breeding for yield where a much larger and complex genetic control is involved.
In such case, a much larger population, and intensive evaluation over years, locations and
environments would be necessary. Ftuther, a much higher estimate of 02A (Additive)
than 02y (Year) , 0'21 (Location), 02d ( Pub.density) and their interactions indicated that
additive variance for these traits was large and selection would be effective and fixable.
Since this model did not provide estimates of dominance or any other interaction
components, therefore, the conclusions based on this may not be very sound. However, in
the self-pollinated crops like soybean, one would anticipate higher additive type of variance
in the advanced generations (F6 or F7). This result supports the findings of Brim and
Cockerham (1961) and Brim (1973), where they reported presence of high additive

variance for these traits in soybean.

5.3.2.2 Additive & Additive x Additive Model
In this model the total variance was partitioned into 25 components without
involving dominance component. The estimates of variances both under full and selected
model provided nearly identical estimates. The estimates of 026 were the same in both,
which indicated effectiveness of the selected model. The estimates of F- values in the
selected model for all the traits indicated that the selected models were adequate to explain

most of the variability in the population. Since the number and type of variance

103

components included in the selected models varied with the characters, this would indicate
that the effects of different variance components differed between traits.The estimates of
components were also slightly readjusted in the selected models.

The estimates of OZAA were higher than 02A in both models for all the characters
except protein. This indicates that AxA type of variances are higher than the additive (A)
type alone for the quantitative traits studied. This finding is in agreement with Hanson et
al.(1967), who reported presence of considerable AxA epistasis in soybean for yield,
maturity, and percent seed coat mottlin g accounting for more than 50% of the total
genotypic variance. Since quantitative traits are controlled by a large number of genes, and
soybean being a fully self-fertilizing crop, these genes should be nearly in a fixed stage.
This indicates that the genes controlling these quantitative traits are mostly in AxA epistasis
or higher order additive types of interactions. These results are in agreement to the

expectations for a self- pollinated crop. There should be more OZAA or higher order
AxA.interactions , and a very low or no 02A. However, these are contradictory to the
findings of Brim and Cockerham (1961), where they reported a high additive variance. It
may be possible that their model was not adequate to separate the effect of AxA from
Additive variance( as we also had similar results in model-1). These information provide a
vital guideline to the breeders of self-pollinated crops. In order for selection to be
effective, there must be adequate AxA or higher order AxA epistatic types of variance in
the population. This will be possible only in the advanced generations. Therefore,
selection for these traits should be delayed until F5 - F6 generation for a greater response.
Early generation selection will not be effective as many of these genes will still be in the
heterozygous condition involving dominance, and will segregate on selfing in the later
generations.
5.3.2.3 Additive and Dominance Model

In this model, estimates of 02A and 62D, and their interactions with other factors

were computed both in full and selected models. Here too, the estimates of oze were the

104

same in both the cases. However, the estimates of 02D were much greater than previously
reported in soybean, while 02A estimates were also much lower. The interaction variances
of dominance(D) with year, location and pubescence density (d) were very high. Such
high estimates of 02D might occur if all the AxD, AxAxD,.or higher order AxD variances
were merged into 02D component in this model. Existence of higher order AxD or very
low 02D has been reported by Gates et al.(1960), Cockerham (1961), Croissant and Tonic
(1971), and Brim (1973). Brim (197 3) reported evidence of heterosis for yield in soybean.
However, it is not yet clear how much is due to 02D, and how much is due to AxD and its
higher type of interactions. If the dominance effects are real and great, and techniques to
produce hybrids are cheap, one might attempt hybrid breeding for heterosis. However,
since the present model does not separate the AxA or AxD effects, the information may not
be very valid.

As is clear now, none of these models included additive, dominance, AxA, AxD
or higher order interactions together. Therefore, none of these models is perfect in itself.
The more factors we include in the model, the more complicated would be the population
structure to develop, and more would be the interaction effects which might jeopardize the
important effects. However, in the present study, the Additive & Additive x
Additive(model-Z) appears to be most suitable for soybean to estimate the components of

variance in F6- F7 generations.

5.4 Estimation of Heritability and Gain from Selection

The estimates of heritability, gain from selection (Gs) and gain from selection as
percent of mean (Gs%) both in Additive, and Additive & Add x Add models appear to be
identical (Tables 24 & 25). However, the estimates were a little higher in Additive model.
The estimates of height, maturity, lodging, seed size and protein were comparatively
higher than those for yield and oil. The heritability of oil was very low which indicated

that selection for this trait will require intensive evaluation over years, locations , and

105

environments. Moreover, a low estimate in this case might have resulted due to little
difference for oil between the parents leading to a lower genotypic variance. Therefore, in
order to breed for high oil, selection of parents with high oil is important to have high
genotypic variance in the population for selection for oil to be effective. However, a trait
with high heritability would indicate low environmental influence, phenotype highly reflect
genotype, and a visual selection would be effective. The Gs % also indicated that response
to selection would be high for height, maturity, lodging and seed size, low for yield, and
poor for protein and oil.

However, in the Additive & Dominance model, because of the low estimates of
62A, only broad-sense heritability were computed which also indicated a similar trend.
The Gs % was very high (>100%) for lodging, height and yield. The heritability(H) for
protein and oil were lower than other characters , and the Gs% were also low. The
heritability (H) calculated from the General model also showed a similar trend: height had a
heritability of 0.99 and those for yield , maturity, lodging, protein and seed size were
around 0.70-0.80, and for oil was 0.46. The heritabilities estimated using parent-
offspring regression (Table 28) were not in agreement with the previous results, but
height again showed the highest value. Yield and lodging had very poor estimates.

It is clear from the results in the present study that the estimates of heritability vary
greatly with the method of estimation. The estimates obtained by Additive or Additive &
Add x Add models are narrow sense, and are more reliable for a self pollinated crop like
soybean. Since these estimates were done in the advanced generations ( F5- F7), and are
based on years and locations evaluation, they should be more precise and reliable.
Usually, the estimates of heritability are higher in the early generation, and decrease in the
later due to decrease of non-additive genetic variance in the later generations (Kelly and
Bliss, 1975). Present findings were in close agreement with Brim (1973), Johnson and
Bemard(1963), and Shannon et al. (1972). Openshaw and Hadley (1984) reported the
heritability of two populations as 0.90 and 0.75 for protein, 0.93 and 0.73 for oil, and

106

0.78 and 0.68 for yield respectively. Their estimates for protein and oil were much higher
than the present findings. This may be due to fact that estimation in earlier generation using
a diverse parental population, or using a different method for estimation might have
resulted in such high estimates of heritability. Parent-offspring regression method of
estimating heritability is effective in absence of dominance and epistasis(Smith and
Kinnman, 1965). Since there is adequate evidence of the presence of epistasis (AxAx...)
for the quantitative traits in soybean, estimates of heritability by parent- offspring
regression in the present study may not be very reliable. The estimates obtained from other
methods were similar. The heritabilities were computed in much later generations based
on year and location testing, they must be very stable and precise as other variance
components like effects of year, location and density, and their interactions have been
separated precisely from the genotypic and phenotypic variances. A moderate to high
estimates of h2 for maturity, height, lodging and seed size would reveal that phenotypic
selection for these traits will be effective. Selection for these traits would not require an
intensive testing of materials over years and locations. However, a moderate estimates for
yield and protein would indicate the influence of higher environmental effects, and the
phenotype would not give a true representation of the genotype, and the visual selection
will not be very effective. Therefore, moderate testing over years and locations would be
necessary for making valid decision. The estimate for yield and oil were usually very low
among all the traits in different models. This may be due to lack of variability for yield and
oil in parents. Alternatively, this may be due to higher environmental influence on them
and phenotypes do not truly represent the genotypes. Hence, for such traits, more testing
over environments would be essential for drawing conclusions. A high environmental
variance may indirectly indicate a comparatively larger number of genes or gene-complex
involved in the inheritance of this trait. Further, more the number of genes involved, more
difficult would be to make progress through the conventional breeding, and a non-

conventional approach (mutation breeding and /or genetic engineering) might be necessary.

107

5.5 Estimation of Correlation Coefficients

The estimation of Pearson's correlation coefficients (Genotypic) and the effects of
year, location, pubescence density, and their interactions on the r—coefficients in the present
study provide an unique and interesting way of understanding the significance of this
estimate in breeding. When the correlations were estimated with the overall or average
(AVE) genotypic means, yield showed a significant positive association with maturity,
seed size and protein. Maturity had a significant positive correlation with all other traits.
However, height exhibited a positive association with maturity and lodging, and significant
negative association with seed size, protein and oil. This indicated that an increase in
height will decrease these traits. Also, lodging had a positive correlation with seed size,
protein and oil, and those with seed size, protein and oil were all positive. These
observations partially support the findings of Johnson and Bernard (1963) where they
found a correlation of 0.4 between yield and maturity, and a very low correlation for yield
and oil. Anand and Torrie (1963) and Kwon and Torrie (1964) also reported a positive
correlation of yield with height, maturity and lodging. However, Byth et al.(1969) found a
negative correlation of yield with height and lodging, which very well supports the current
findings. They also found a higher correlation of yield with protein than with oil.
Hartwig and Edwards (1970) also found a negative association of yield and height.
Johnson et al.(1955) reported a very strong and positive correlation between yield and seed
size which fully supports our findings. However, a high positive association for protein
and oil on the AVE contradicted the earlier findings (Shorter et al.,1976; Brim and Burton,
1979; Bmton and Brim,1983). This might have arisen due to the balancing of positive and
negative effects when averaged over years, locations and density.

In this study, the correlation coefficients were further computed across year,
location, pubescence density and their interactions. In many cases, the correlations showed
a stable response, while in others the estimates and even the direction of effect (plus or

minus) changed drastically. For example, when year effects were compared, the

108

association between yield and maturity was negative in 1989, and positive in 1990.
Likewise, when the correlations were computed by location or pubescence density, they
were in agreement of the ovrall (AVE) estimates. Again, when these estimates were
further computed by location within the year, they were stable for some and quite variable
for others. The association of yield with maturity and oil were almost negative. The
estimates for seed size with protein and oil were also negative contradictory to the overall
‘ (AVE) estimates. Further, the effects of density within the year produced the identical
results. However, when the correlations were computed by locations within the year, the
estimates greatly supported the overall estimates. Therefore, a more detailed computation
of correlation was done by pubescence density within location and year (Table 33) as
compared with the overall (AVE) estimates. It is evident from the table that there is a
tendency of negative association of yield with manuity, height and oil, while seed size and
protein showed a significant positive association with yield. However, there was no
association of yield with lodging. It may be possible that lodging occured at late maturity
without affecting the yield. There was a strong positive correlation of maturity with height
and lodging, while it approached negative with seed size, protein and oil. Height and
lodging had a strong positive association. The associations between lodging and seed size
were positive for some and negative for others. It was interesting to see that correlation of
protein and oil were very high and negative in all cases, while strongly positive on overall
(AVE) basis. This negative association of protein and oil supported the findings of earlier
workers.

The information on correlation coefficients elucidate the situation on how they are
influenced by factors like year, location, pubescence density and their interactions, and
methods of estimation. It is evident that some of these estimates are highly stable across
the factors effects, while others change considerably, and may produce misleading results.
Therefore, those which are quite stable, eg. positive association of yield with seed size and

protein, of maturity with height and lodging, of height with lodging, are well established,

109

and should be used in breeding. The information on correlation can facilitate breeders in
utilizing indirect selection for quantitative traits with low heritability, or for traits which
measurements are difficult and expensive. However, for others the estimates are
influenced by other environmental factors like year, location, pubescence density and their
interactions. In such cases, it would be necessary to evaluate the materials over more years
and locations, and over a range of environments in order to make any reliable inferences.
Based on our detailed study, we can clearly show that there is a strong negative correlation
between protein and oil. The estimates of correlation coefficients are important to breeders
in deciding breeding procedures and selection programs. A high positive or negative
correlation of the secondary trait(s) with the primary trait could be used for indirect
selection of the primary trait. Selection for high yield will increase seed size and protein,
but decrease oil content. A strong negative correlation between protein and oil will indicate
that selection for one will decrease the other. In situations where we need to increase both
protein and oil, first we should screen more germplasm to find parents with high protein
and oil, or use a different breeding technique like recurrent selection, mutation breeding

and] or genetic engineering to generate more variability.

5.6 Estimation of Regression Coefficients

The estimation of regression coefficients and developing the prediction equations
are still another way of looking at the dependence of primary traits on the secondary or
independent traits. In the stepwise regression, unimportant variables were dropped. For
example, for predicting yield, the model included maturity, lodging , seed size and oil.
Similarly, for protein the model included maturity, seed size and oil, while for oil, variables
yield, maturity, height, seed size and protein were included. The estimates and direction of
regression coefficients will indicate the relationship of independent variables with the
dependent variable. The computation of the prediction equation or response curve will

indicate the degree of change in the dependent variable with a change in one or more

110

independent variables. This information is crucial in breeding programs for predicting
response to selection. Such estimates are of greater significance particularly when the
measurements of dependent variable is difficult, time taking, or need special equipment and
facility. For example, these prediction equations could be used to predict gain in yield
protein or oil without their immediate analysis. This will help breeders to cut time and

resources, and increase efficiency.

6 SUMMARY AND CONCLUSIONS

Most of the cultivated varieties of soybean have a dense covering of erect
pubescence (hairs) on stem, leaf, calyx and pod. Considerable genetic variation in
pubescence size, form, density, durability and color has been found in the germplasm
collection. Pubescence density is a simple inherited trait controlled by a single dominant or
recessive gene. The trait has been found to affect plant Vigor, and resistance to insects and
drought. Pubescence density has been found to increase height and lodging , and reduce
yield in the indeterminate cultivars. These information, however, are based on a few
genotypes. There is inadequate information on how dense pubescence affect the
quantitative traits of economic importance, like yield, maturity, height, lodging, seed size,
protein and oil, and their genetic components of variance, heritability and correlations. A
clear understanding of these underlying genetic principles is crucial to effective breeding
program.

The present research was designed i) to study the effect of dense pubescence on
these quantitative traits, ii) to estimate genetic components of variance, iii) to estimate g x e
interactions, iv) to determine heritability and gain from selection, V) to compute correlations
and regression coefficients among these traits, and Vi) to discuss the implications of
findings to soybean breeding.

The test lines were developed from the cross of Wells H (++) with normal
pubescence and Harosoy (Pde) with dense pubescence. Dense and normal lines were
developed in a nested design from F2 and F3 (F2) derived progenies. Sixty progeny lines
with the two parents were evaluated in F6 and F7 generations for two years at two
locations in Michigan.

The comparison of means and simple statistics for different traits revealed that they
were highly influenced by year, location, pubescence density, and their interactions.

Pubescence density increased maturity, height, lodging and oil, and reduced yield It
1 11

112

exhibited no significant effect on seed size and protein. However, when the effects of
pubescence density were studied within year and location, only height had a stable
response, while lodging and oil exhibited response in only one year. Therefore, it is
important to evaluate materials over years and locations for making reliable conclusions
about the effects of pubescence density on quantitative traits.

The analysis of variance indicated a significant effect of genotypes on all the traits.
Therefore, it may be possible to select lines of interest in the population. Pubescence
density (D) did not show any main effect on any trait, however, the interactions were
present, and varied with the traits. Therefore, it is important to determine the genotypes,
locations and environments (drought), where response is evident, and use dense
pubescence as an important trait in breeding. Since breeding of quantitative traits is
complex and time taking due to polygenic inheritance, a strong linkage of these traits with
pubescence density (simple inheritance) may be employed for indirect selection.

The genetic components of variance differed with the models used. In general,
Additive and Additive x Additive model appeared most suitable in this study. The
estimates of variance Add x Add were in general higher than Additive variance for most
traits. The variance due to pubescence density (d) and its interactions with other factors
were much lower than the genetic components for all the traits except height. This indicates
that pubescence density does not have a major effect on the genetic components of variance
of these quantitative traits, and so as on the estimates of heritability. Year and location
effects were also evident and varied with the traits. The heritability estimates computed
under Additive, and Add & AddxAdd models were nearly similar. The estimates were high
for height, maturity and lodging (>0.50) , moderate for seed size and yield (0.35-0.49)
and low for protein and oil (0.25-0.34). However, in the General model ANOVA the
estimates of heritability (H) as they were broad-sense, were generally high. The traits with
high heritability indicate that the phenotype highly reflect the genotype, and visual selection

may be effective. For the traits with low heritability evaluation over years and locations

113

may be necessary for making conclusions. Pubescence density did not show much
influence on heritability of any trait except height. Alternatively, isozyme and RFLP
techniques may be helpful for indirect selection, provided they are shown to be associated
with useful agronomic traits.

.The correlation coefficients were highly influenced by year, location, pubescence
density and their interactions. The conclusions on correlations made using overall
genotypic means may be misleading in some cases. We computed the correlations using
the main and interaction means of year, location and pubescence density. Some of the
coefficients, like yield with seed size and protein, maturity with height and lodging, and
height with lodging were consistently high and positive, while others differed across these
factors. The association between protein and oil was significant and positive when
computed using overall means, but exhibited significant negative when computed by above
factors. Therefore, it is important to use the traits which show a stable correlation for
indirect selection in breeding. The general and stepwise regression coefficients were
computed and used to predict response of yield, protein and oil. Based on these, maturity,
lodging, seed size and oil were important variables for predicting yield; maturity, seed size
and oil for protein; and all except lodging were important for oil. They can be used for
predicting responses to selection for these traits in breeding. Such estimates are of greater
significance particularly when measurements of dependent variable is difficult, time taking,
or needs special equipment and facilities.

BIBLIOGRAPHY

BIBLIOGRAPHY

Anand, S. C., and J. H. Torrie. 1963. Heritability of yield and other traits and inter
correlationships among traits in the F3 and F4 generations of three soybean

crosses. Crop Sci. 3:508-511.

Anderson, W. F., C. C. Holbrook, and J. C. Wynne. 1991. Heritability and early
generation selection for resistance to early and late leaf spot in peanut. Crop Sci.
31:588-593.

Baldocchi, D. D., S. B. Verrna, N. J. Rosenberg, B. L. Blad, A. Gary, and J. E. Spech.
1983. Leaf pubescence effects on mass and energy exchange between soybean
canopies and the atmosphere. Agron. J. 75:537-542.

Bernard, R. L. 1975. The inheritance of near-gray pubescence color.
Soybean Genet. Newsl. 2:31-33.

Bernard, R. L., and B. B. Singh. 1969. Inheritance of pubescence type in soybeans:
glabrous, curly, dense, sparse, and puberulent. Crop Sci. 9:192-197.

Bernard, R. L., and M. G. Weiss. 1973. Qualitative genetics. p.117-146. In B. E.
Caldwell (ed.) Soybeans. American Society of Agronomy, Inc. Pub. Maidison,
WI,U.S.A. p-681.

Brim, C. A. 1973. Quantitative genetics and breeding. p. 155-183.
In B. E. Caldwell (ed.) Soybeans. American Society of Agronomy, Inc. Pub.
Maidison, WI,U.S.A. p-68l.

Brim, C. A., and C. C. Cockerham. 1961. Inheritance of quantitative characters in
soybeans. Crop Sci. 1:187-190.

Brim, C. A., and J. W. Burton. 1979. Recurrent selection in soybeans. II. Selection for
increased percent protein in seeds. Crop Sci. 19:494-498.

Broersma, D. B., R. L. Bernard, and W. H. Luckrnann. 1972. Some effects of soybean
pubescence on population of potato leafhopper. J.Econ. Ent. 65:78-82.

Burton, J. W. 1987. Quantitative genetics: Results relevant to soybean breeding. p.211-

247. In J. R. Wilcox (ed.) Soybeans: Improvement, production, and uses.
Agronomy, Inc. Pub. Madison, WI, U.S.A.

114

115

Burton, J. W., A. E. Purcell, and W. M. Walter, Jr. 1982. Methionine concentration in
soybean protein from populations selected for increased percent protein. Crop Sci.
22:430-432.

Burton, J. W., and C. A. Brim. 1981. Recurrent selection in soybeans. III. Selection for
increased percent oil in seeds. Crop Sci. 21:31-34.

Burton, J. W., R. F. Wilcox, and C. A. Brim. 1983. Recurrent selection in soybeans. IV.
Selection for increased oleic acid percentage in seed oil. Crop Sci. 23:744—747.

Buttery, B. R., and R. I. Buzzell. 1973. Varietal differences in leaf flavonoids of
soybeans. Crop Sci. 13:103-106.

Buttery, B. R., R. I. Buzzell, and W. I. Findlay. 1981. Relationships among
photosynthetic rate, bean yield and other characters in field grown cultivars of
soybean. Can. J. Plant Sci. 61:191-198.

Buzzell, R. I., and B. R. Buttery. 1977. Soybean harvest index in hill-plots.Crop Sci.
17:968-970.

Byth, D. E., and B. E. Caldwell. 1969. Correlated truncation selection for yield in
soybeans. Crop Sci. 9:699-702.

Byth, D. E., B. E. Caldwell, and C. R. Weber. 1969. Specific and non-specific index
selection in soybeans, Glycine max L.(Merrill). Crop Sci. 9:702-705.

Caldwell, B. E. 1973. (ed.) Soybeans. p.ix-x. American Society of Agronomy, Inc. Pub.
Madison, WI, USA. p. 681.

Carlson, J. B. 1973. Morphology. p.17-66. In B. E. Caldwell (ed.) Soybeans. American
Society of Agronomy, Inc. Madison, WI, U.S.A.

Cockerham,C. C. 1963. Estimation of genetic variances. p.53-94. In
W. D. Hanson and H. F. Robinson (ed.) Statistical genetics and plant breeding.
National Academy of Sciences- National Research Council, Washington, DC.

Cockerham, C. C. 1983. Covariances of relatives from self-fertilization. Crop Sci.
23:1177-1180.

Cooper, R. L., and A. Waranyuwat. 1985. Effects of three genes (Pd, km, and 1n) on
plant height, lodging, and seed yield in indeterminate and determinate near isogenic
lines of soybeans. Crop Sci. 25:90-92.

 

Croissant, C. L., and J. H. Torrie. 1971. Evidence of non-additive effects and linkages in
two hybrid populations of soybeans. Crop Sci. 11:675-677.

Dzikowski, B. 1937. Studia nad soja Glycine hispida M. Maxim. C2. Anatomia. Pametnik
Panstwowego Instytutu Naukowego Gospodarstwa Wiejskiego w Pulawach. Tom
XVI. Rosprawa Nr. 258:229-265.

Ecochard, R. and Y. Ravelomanantsoa. 1982. Genetic correlations derived from full-sib
relationships in soybean (Glycine max L. Merr.) Theor. Appl. Genet. 63:9-15.

116

Ehleringer, J. R., and H. A. Mooney. 1978. Leaf hairs: Effects on physiological activity
and adaptive value to a desert shrub. Oecologia 37:183-200.

Falconer, D. S. 1980. Introduction to quantitative genetics. The Ronald Press Co., NY.
FAO. 1989. Soybean. p.155-156. In FAO Production year book. FAO, Rome.

Fehr, W. R. 1987. Heritability. In Priciples of cultivar development: Theory and
techniques. Vol. 1:95-96. McGraw-I-Iill, Inc. NY.

Ford, D. M., R. Shibles, and D. E. Green. 1983. Growth and yield of soybean lines
selected for divergent leaf photosynthetic ability. Crop Sci. 23:517-520.

Franceschi, V. R., and R. T. Giaquinta. 1983. Glandular trichomes of soybean leaves:
Cytological differentiation from initiation through senescence. Bot. Gaz.
144: 175-1 84.

Gates, C. E., C. R. Weber, and T. W. Homer. 1960. A linkage study of quantitative
characters in asoybean cross. Agron J. 52:45-49.

Gausman, H. W., and R. Cardenas. 1973. Light reflectance by leaflets of pubescent,
normal and glabrous lines of soybeans. Agron. J. 65:837-838.

Ghorashy, S. R., J. W. Pendelton, R. L. Bernard, and M. E. Bauer. 1971. Effects of leaf
pubescence on transpiration, photosynthetic rate and seed yield of three near-
isogenic lines of soybeans. Crop Sci. 11:426-427.

Hanson, W. D., A. H. Probst, and B. E. Caldwell. 1967. Evaluadon of a population of
soybean genotypes with implication for improving self-pollinated crops. Crop Sci.
7:99-103.

Hanson, W. D., R. C. Leffel, and R. W. Howell. 1961. Genetic analysis of energy
production in the soybean. Crop Sci. 1:121-126.

Hartung, R. C., J. E. Spech, and J. W. Williams. 1980. Agronomic performance of
selected soybean morphological variants in irrigation culttue with two row
spacings. Crop Sci. 20:604-609.

Hartwig, E. E., and C. J. Edwarder. 1970. Effects of morphological characteristics upon
seed yield in soybeans. Agron. J. 62: 64-65.

Hillsman, K. J., and H. W. Carter. 1981. Performance of F1 hybrid soybeans in
replicated row trials. p.63. Agron. Abstr., American Society of Agronomy,
Madison,WI, U.S.A.

Johnson, H. W., and E. A. Hollowell. 1935. Pubescent and glabrous characters of
soybeans as related to resistance to injm'y by the potato leaflropper. J. Agri. Res.
51 :371-38 1.

Johnson, H. W., H. F. Robinson, and R. E. Comstock. 1955. Genotypic and phenotypic
correlations in soybeans and their implications in selection. Agron. J. 47:477-483.

117

Johnson, H. W., and R. L. Bernard. 1963. Soybean genetics and breeding. p.1-73. In
A.G. Norman(ed.) The soybean.
Academic Press, NY.

Kawahara, E. 1963. Studies on the gene analysis of soybeans (English summary). Tohoku
Agric. Exp. Sta. Acad. Rep. 26:79-147.

Kelly, J. D., and F. A. Bliss. 1975. Heritability estimates of percentage seed protein and
available methionine and correlations with yield in drybeans. Crop Sci.
15:753-757.

Kwon, S. H., and J. H. Torrie. 1964. Heritability of and interrelationships among traits of
two soybean populations. Crop Sci. 4:196-198.

Lush, J. L. 1940. Intrasire correlations or regressions of offspring on dam as a method of
estimating heritability of characteristics. Proc. Am. Soc. An. Prod. :293-301.

Morse, W. J ., and J. L. Carter. 1937. Improvement in soybeans. p.1154-1189 Year book
Agr.(USDA), US Govt. printing office, Washington.

Nagai, I., and S. Saito. 1923. Linked factors in soybeans. Jap. J. Bot. 1:121-136.

Nelson, R. L., and R. L. Bernard. 1984. Production and performance of hybrid soybeans.
Crop Sci. 24:549-553.

NGLIP. 1988. Annual Report. National Grain Legume Improvement Program(NGLIP),
Rampur, Nepal.

Openshaw, S. J ., and H. H. Hadley. 1981. Selection to modify sugar content of soybean
seeds. Crop Sci. 21:805-808.

Openshaw, S. J ., and H. H. Hadley. 1984. Selection indexes to modify protein
concentration of soybean seeds. Crop Sci. 24:1-4.

Owen, F. V. 1927. Inheritance studies in soybeans. II. Glabrousness, color of
pubescence, time to maturity, and linkage relations. Genetics 12:519-529.

Palmer, R. G., and T. C. Kilen. 1987. Qualitative genetics and cytogenetics. p.135-209.
In J .R. Wilcox(ed.) Soybeans: Improvement, production, and uses. Agronomy,
Inc. Pub. Madison, WI, U.S.A.

Pandy, J. P., and J. H. Torrie. 1973. Path coefficient analysis of seed yield '
components in soybeans (Glycine max (L.) Merr.). Crop Sci. 13:505-507.

Piper, C. V., and W. J. Morse. 1923. The soybeans. McGraw Hill Book Co.,NY. p.329.
Poos, F. W. 1929. Leaflropper injury to legumes. J. Econ. Ent. 22:146-153.
Poos, F. W., and F. F. Smith. 1931. A comparison of oviposition and nymphal

development of Empoascafabae (Harris) on different host plants. J. Econ. Ent.
24:361-37 1.

118

Powell, W., P. D. S. Caligari, W. T. B. Thomas, and J. L. Jinks. 1985. The effects of
major genes on quantitatively varying characters in barley: 2. The denso and
daylength response loci. Heredity 54:349-352.

Powers, L. 1955. Components of variance method and partitioning method of genetic
analysis applied to weight per unit of tomato hybrid and parental populations.
USDA Tech. Bull. 1131. U.S. Govt. Printing Office, Washington, DC.

Probst, A. H., and R. W. Judd. 1973. Origin, U.S. history and development, and world
distribution. p.1-12. In B.E. Caldwell (ed.) Soybeans: American Society of
Agronomy, Inc. Pub. Madison, WI, U.S.A.

Rasmusson, D. C. and R. L. Glass. 1967 . Estimates of genetic and envionmental
variability in barley. Crop Sci. 7:185-188.

SAS. 1985. SAS User's Guide: Statistics. Version 5 ed. SAS Ins. Inc. Cary, NC.

Shannon, J. G., J. R. Wilcox, and A. H. Probst. 1972. Estimated gains from selection
from protein and yield in the F4 generation of six soybean populations. Crop Sci.
12:824-826.

Shorter, R., D. E. Byth, and V. E. Mungomery. 1976. Estimates of selection parameters
associated with protein and oil content of soybean seeds (Glycine max (L.) Merr.).
Aust. J. Agric. Res. 28:211-222.

Simpson, A. M. Jr., and J. R. Wilcox. 1983. Genetic and phenotypic associations of
agronomic characteristics in four high protein soybean populations. Crop Sci.
23:1077-1081.

Singh, B. B., H. H. Hadley, and R. L. Bernard. 1971. Morphology of pubescence in
soybeans and its relationship to plant vigor. Crop Sci. 11:13-16.

Smith, J. D., and M. L. Kinman. 1965. The use of parent offspring regression as an
estimator of heritability. Crop Sci. 5:595-596.

Specht, J. E., J. H. Williams, and D. R. Pearson. 1985. Near-isogenic analysis of
soybean pubescence genes. Crop Sci. 25:92-96.

Sprague, G. F., and W. T. Federer. 1951. A comparison of variance components in corn
yield trials. Agron. J. 43:535-541.

Takahashi, Y., and J. Fukuyama. 1919. Morphological and genetic studies on the soybean
(In Japanese). Hokkaido Agr.Exp. Sta. Rep. 10.

USDA. 1990. Agricultural Statistics, 1989.

Van de Logt, M. B., M. A. Brinkman, and R. A. Forsberg. 1984. Variance components
and linkage effects for qualitatively inherited traits in two barley crosses. Crop Sci.
24:323-327.

Waggoner, P. E. 1966. Decreasing transpiration and the effect upon growth. p.49-68.In
Plant environment and efficient water use (Monograph). American Society of
Agronomy, Madison, WI, U.S.A.

119

Weber, C. R., L. T. Empig, and J. C. Thorne. 1970. Heterotic performance and
combining ability of two-way F1 soybean hybrids. Crop Sci. 10: 159-160.

Williams, L. E. 1950. Structure and genetic characteristics of the soybeans. p.111-134. In
K. S. Markley (ed.) Soybean and soybean products. Vol. I. Interscience Pub.
NY.

Wolfenbarger, D. A., and J. P. Sleesman. 1963. Variation in susceptibility of soybean
pubescent types, broad bean, and runner bean varieties and plant introductions to
the potato leaflropper. J. Econ. Ent. 56:895-897.

Woodworth, C. M. 1921. Inheritance of cotyledon, seed coat, hilum, and pubescence
color in soybeans. Genetics 6:487-553.

Woodworth, C. M., and C. Veatch. 1929. Inheritance of pubescence in soybeans and its
relation to pod color. Genetics 14:512-518.

Woolley, J. T. 1964. Water relations of soybean leaf hairs. Agron. J. 56:569-571.
Yiran, Y., T. Byme, and J. Harding. 1990. Quantitative genetic analysis of flowering time

in the Davis population of gerbera. 1. Components of genetic variance and
heritability. Crop Sci. 30:19-23.

APPENDICES

Appendix A.l. Overall Means of the Genotypes across years and

 

 

locations.

Entry #Acc.# Source Pub Yield Mat HGT LDG Size Protein oil

1. E82026 Wells II N 3.74 27.3 88.1 1.0 17.5 38.9 19.5
2. 183170 HSY-PD D 3.11 28.1 103.1 2.1 20.4 39.4 19.3
3. E89001 01-1 + N 3.00 36.7 114.4 3.0 18.7 38.2 19.1
4. E89002 01-1 PD D 2.95 37.8 132.7 3.6 19.2 38.1 18.6
5. E89003 01-2 + N 3.47 37.0 108.8 2.8 19.0 37.7 19.5
6. E89004 01-2 PD D 2.97 37 .0 125.5 3.5 18.6 38.4 19.1
7. E89005 02-1 + N 4.06 30.1 100.1 2.0 18.8 38.9 19.5
8. E89006 02-1 PD D 3.64 29.8 102.7 2.5 20.1 39.6 19.1
9. E89007 02-2 + N 3.61 28.8 102.0 2.8 18.4 39.2 19.3
10. E89008 02-2 PD D 3.52 30.7 113.8 3.0 19.1 39.5 19.1
11. E89009 04-1 + N 3.91 30.1 90.5 1.8 19.4 39.4 18.9
12. E89010 04-1 PD D 3.39 31.0 109.0 2.2 20.0 39.6 19.3
13. E89011 04-2 + N 3.61 29.1 94.0 2.3 19.8 39.3 19.4
14. E89012 04-2 PD D 3.30 30.0 106.1 2.6 20.2 39.6 19.4
15. E89013 05-1 + N 3.74 31.1 103.2 2.3 21.7 39.1 19.5
16. E89014 05-1 PD D 3.68 30.1 101.3 2.0 21.5 39.0 19.2
17. E89015 05-2 + N 3.11 29.0 89.2 2.3 19.9 38.8 19.4
18. E89016 05-2 PD D 3.01 28.7 102.0 2.5 20.7 39.5 18.8
19. E89017 06-1 + N 3.49 27.7 89.7 2.0 21.1 39.4 19.2
20. E89018 06-1 PD D 3.07 39.2 118.0 3.0 17.7 38.3 19.4
21. E89019 062 + N 3.40 28.1 92.6 1.7 20.3 39.7 19.1
22. E89020 062 PD D 3.52 28.2 96.0 2.5 21.2 39.2 19.1
23. E89021 07-1 + N 3.62 27.6 98.1 1.7 19.3 39.8 19.1
24. E89022 07-1 PD D 3.37 28.6 102.1 2.0 20.4 39.7 19.1
25. E89023 07-2 + N 3.87 29.5 105.1 2.2 18.2 38.9 19.3
26. E89024 07-2 PD D 3.59 30.6 120.6 2.5 18.6 39.8 18.8
27. E89025 08-1 + N 3.55 28.8 102.5 2.0 18.2 39.7 19.2
28. E89026 08-1 PD D 3.29 30.7 112.0 2.3 18.8 40.0 18.9
29. E89027 08-2 + N 3.96 28.6 101.3 1.3 17.4 39.7 18.9
30. E89028 082 PD D 3.34 30.1 116.8 2.2 18.4 39.7 18.8
31. E89029 09-1 + N 2.73 27.0 87.3 1.3 21.9 39.4 19.1
32. E89030 09-1 PD D 2.79 38.2 136.2 3.5 17.4 38.3 19.1
33. E89031 09-2 + N 2.99 38.5 115.6 2.8 17.7 38.7 19.1

 

Appendix A.1.

Continued....

 

 

Entry #Acc.# Source Pub Yield Mat HGT LDG Size Protein oil

34. E89032 09-2 PD D 2.84 38.8 118.8 3.0 17.4 38.3 18.8
35. E89033 101 + N 3.59 27.8 97.8 1.1 18.1 39.2 19.0
36. E89034 101 PD D 3.44 28.8 106.5 1.8 18.7 39.4 18.9
37. E89035 10-2 + N 3.46 27.8 82.7 1.1 18.4 39.1 19.1
38. E89036 10—2 PD D 3.25 28.2 91.7 1.1 19.3 39.4 18.7
39. E89037 11-1 + N 3.67 28.2 94.1 1.5 19.3 39.4 19.1
40. E89038 11-1 PD D 3.69 28.3 101.5 2.2 19.3 39.7 19.2
41. E89039 11-2 + N 3.68 24.8 91.6 1.0 18.2 39.0 19.7
42. E89040 11-2 PD D 3.36 26.6 97.6 1.3 20.1 39.1 19.5
43. E89041 14-1 + N 3.64 28.3 92.3 2.3 18.5 39.0 19.2
44. E89042 14-1 PD D 3.64 28.6 105.8 2.3 18.8 39.5 18.9
45. E89043 14-2 + N 3.34 26.5 87.7 1.6 19.1 38.5 19.9
46. E89044 14-2 PD D 2.91 28.3 103.8 2.2 20.5 38.5 19.9
47. E89045 15-1 + N 3.85 30.0 101.8 2.5 20.3 40.6 18.7
48. E89046 15-1 PD D 3.27 30.3 109.3 2.7 19.6 40.5 18.6
49. E89047 15-2 + N 3.53 28.8 101.7 2.5 19.6 40.0 18.8
50. E89048 15-2 PD D 3.31 28.7 109.3 2.6 19.6 39.8 18.9
51. E89049 17-1 + N 2.59 27.5 99.7 1.5 18.5 37.6 19.7
52. E89050 17-1 PD D 2.89 30.2 108.0 1.8 18.8 38.3 19.7
53. E89051 17-2 «1» N 3.36 34.6 114.7 2.7 18.7 38.4 19.9
54. E89052 17-2 PD D 3.21 35.3 115.5 3.0 18.7 38.8 19.2
55. E89053 18-1 + N 3.04 26.8 86.0 1.5 18.4 39.2 19.4
56. E89054 18-1 PD D 2.89 36.1 121.3 3.3 17.7 38.0 20.1
57. E89055 18-2 + N 2.97 28.1 78.0 1.4 19.3 38.6 19.8
58. E89056 18-2 PD D 3.31 29.0 92.0 1.5 20.1 39.0 19.3
59. E89057 19—1 + N 3.48 28.5 96.0 2.5 19.1 39.5 19.0
60. E89058 19-1 PD D 3.23 30.1 104.6 2.7 19.5 39.7 18.9
61. E89059 19-2 + N 2.98 25.7 85.3 1.5 17.8 39.0 19.4
62. E89060 19-2 PD D 2.98 26.8 97.0 1.7 19.0 38.4 19.6

 

Where, Pub. = Pubescence type, Yield = Grain yield (t/ha), Mat = Maturity in days
(after 8/31), HGT = Plant height (cm), LDG = Lodging in 1-4(1 resistant, 4 susceptible)
Size = Seed size (g/100 seeds), Protein = % protein in seed, Oil = % oil in seed,

D = Dense pubescence, and N = Normal pubescence.

Appendix B.l.

122

Analysis of variance table of the General model

 

 

 

for Yield.
# Source df SS MS MSe dfe F Sig.
1 . Year(Y) 1 6.03 6.03 3.81 1 1.59 ns
2. Location(L) 1 82.02 82.02 3.75 1 21.84 as
3. Y x L 1 3.72 3.72 0.32 61 11.50 **
4. Rep(Y x L) 4 10.83 2.71 0.20 244 13.55 **
5. Genotype(G) 61 54.65 0.89 0.43 49 2.07 **
6. Y x G 61 24.63 0.40 0.32 61 1.25 ns
7. L x G 61 21.53 0.35 0.32 61 1.09 ns
8. Y x L x G 61 19.77 0.32 0.20 244 1.62 **
9. Error 244 48.80 0.20
C.V. = 13.31% Mean = 3.35 t/ha

LSD(0.05) = 0.87t/ha

ns = non-significant.

LSD(0.01) = 1.15t/ha
*, **= Significant at p>0.05 and p>0.01, respectively.

Appendix B.2. Analysis of variance table of the General model

for Maturity.

 

 

 

# Source df SS MS MSe dfe F Sig.
1 . Year(Y) 1 4931.6 4931.6 2774.6 1 1.77 ns
2. Location(L) 1 2620.2 2620.2 2769.1 1 0.95 ns
3. Y x L 1 2729.3 2729.3 9.6 61 288.47 **
4. Rep(YL) 4 114.8 28.7 3.1 244 9.31 **
5. Genotype(G) 61 6165.5 101.1 14.7 33 6.87 **
6. Y x G 61 912.1 15.0 9.6 61 1.55 *
7 L x G 61 570.9 9.4 9.6 61 0.97 ns
8. Y x L x G 61 585.8 9.6 3.1 244 3.11 **
9. Error 244 752.2 3.1

C.V. = 5.81% Mean= 30.2 days after 8/31

LSD(0.05) =3 .44 days

as = non-significant.

LSD(0.01) =4.52 days
*, **= Significant at p>0.05 and p>0.01, respectively.

123

Appendix B.3. Analysis of variance table of the General model

 

 

 

for Height.

# Source df SS MS MSe dfe F Sig
1. Year(Y) 1 285.0 285 .0 1197.6 1 0.23 ns
2. Location(L) 1 8210.3 8210.3 1124.9 1 7.29 ns
3. Y x L 1 1128.0 1128.0 68.0 61 16.58 **
4. Rep(YL) 4 10017.2 2504.3 69.5 244 36.03 **
5. Genotype(G) 61 69385.5 1137.4 134.4 40 8.46 **
6. Y x G 61 8392.1 137.6 68.0 61 2.02 **
7 L x G 61 3957.7 64.9 68.0 61 0.95 ns
8. Y x L x G 61 4148.6 68.0 69.5 244 0.98 ns
9. Error 244 16960.8 69.5

C.V.=8.10% Mean = 102.8 cm

LSD(0.05) =16.34 cm

as = non-significant.

LSD(0.01) = 21.45 cm
*, **= Significant at p>0.05 and p>0.01, respectively.

Appendix B.4. Analysis of variance table of the General model

for Lodging.

 

 

 

# Source df SS MS MSe dfe F Sig.
1. Year(Y) 1 44.8 44.8 0.04 1 1278.00 *
2. Location(L) 1 8.0 8.0 0.29 8 27.68 **
3. Y x L 1 0.0 0.0 0.42 61 0.01 ns
4. Rep(YL) 4 31.8 9.7 0.34 244 28.47 **
5. Genotype(G) 61 212.8 3.5 0.74 38 4.74 **
6. Y x G 61 27.4 0.5 0.42 61 1.07 ns
7 L x G 61 42.9 0.7 0.42 61 1.68 *
8. Y x L x G 61 25.4 0.4 0.34 244 1.22 *
9. Error 244 83.3 0.3

C.V. = 26.51% Mean = 2.2

LSD(0.05) = 1.14

LSD(0.01) = 1.50

*, **= Significant at p>0.05 and p>0.01, respectively.

ns = non-significant.

Appendix B.5. Analysis of variance table of the General model
for Seed size.

124

 

 

 

# Source df SS MS MSe dfe F Sig.
1. Year(Y) 1 1379.1 1379.1 68.2 1 20.20 ns
2. Location(L) 1 73 .4 7 3.4 68 .6 1 1 .07 ns
3. Y x L 1 69.8 69.8 2.6 61 27.10 **
4. Rep(YL) 4 91.4 22.8 0.6 244 36.90 **
5. Genotype(G) 61 586.6 9.6 2.6 61 3.74 **
6. Y x G 61 58.9 1.0 2.6 61 0.37 ns
7 L x G 61 83.8 1.4 2.6 61 0.53 ns
8. Y x L x G 61 157.3 2.6 0.6 244 4.17 **
9. Error 244 150.9 0.6

C.V. = 4.08% Mean = 19.23 gm

LSD(0.05) = 1.54gm

LSD(0.01) = 2.02gm

*, **= Significant at p>0.05 and p>0.01, respectively.

ns = non-significant.

Appendix B.6. Analysis of variance table of the General model

for Protein.

 

 

 

# Source df SS MS MSe dfe F Sig.
1. Year(Y) 1 5628.1 5628.10 8.75 1 643.20 *
2. Location(L) 1 21.9 21.97 8.96 1 2.45 ns
3. Y x L 1 8.5 8.57 0.53 61 16.10 **
4. Rep(YL) 4 0.4 0.1 1 0.006 244 1 .87 ns
5. Genotype(G) 61 195.3 3.20 1.10 45 2.91 **
6. Y x G 61 43.4 0.71 0.53 61 1.38 *
7 L x G 61 56.5 0.92 0.53 61 1.74 *
8. Y x L x G 61 32.5 0.53 0.006 244 88.30 *"‘
9. Error 244 1.4 0.006

C. V. = 0.19 % Mean = 39.16 %

LSD(0.05) = 0.15 %

LSD(0.01) = 0.19 %

*, **= Significant at p>0.05 and p>0.01, respectively.

ns = non-significant.

125

Appendix B.7. Analysis of variance table of the General model

 

 

 

for Oil.
# Source df SS MS MSe dfe F Sig.
1. Year(Y) 1 130.2 130.20 0.68 3 191.40 **
2. Location(L) 1 0.3 0.26 0.45 1 0.59 ns
3. Y x L 1 0.4 0.41 0.24 61 1.72 ns
4. Rep(YL) 4 0.01 0.003 0.006 244 0.42 ns
5. Genotype(G) 61 57.3 0.94 0.54 46 1.72 *
6. Y x G 61 30.9 0.51 0.24 61 2.13 **
7 L x G 61 16.7 0.27 0.24 61 1.14 ns
8. Y x L x G 61 14.6 0.24 0.006 244 40.00 **
9. Error 244 1.5 0.006
C.V. = 0.40 % Mean = 19.24 %
LSD(0.05) = 0.15 % LSD(0.01) = 0.20 %

*, **= Significant at p>0.05 and p>0.01, respectively.
ns = non-significant.

126

Appendix C.l. Analysis of variance table of the Nested model

for Yield.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# Sorn'ce df SS MS MSe dfe F
1. Year(Y) 1 6.323 6.323 3.746 1 1.68
2. Location(L) 1 76.501 76.501 3.478 1 21.99
3. YxL 1 3.413 3.413 2.882 5 1.18
4. Rep(Y,L) 4 10.173 2.543 0.205 236 12.36**
5. F2 14 28.292 2.021 1.475 11 1.37
6. F3(F2) 15 12.883 0.858 0.326 3 2.63
7. Puh.Density(D) 1 4.193 4.193 1.246 0 3.36
8. F2 x D 14 3.576 0.255 0.064 0 3.99
9. F3(F2) x D 15 3.988 0.265 0.316 15 0.84
10. L x F2 14 8.537 0.609 0.552 6 1.11
11. L x F3(F2) 15 6.704 0.447 0.440 15 1.01
12. L x D 1 0.834 0.834 0.943 0 0.88
13. Y x F2 14 12.29 0.878 0.319 15 274*
14. Y x F3(F2) 15 4.794 0.319 0.545 14 0.58
15. Y x D 1 0.961 0.961 0.991 0 0.97
16. Y x F2 x D 14 2.436 0.174 0.341 16 0.51
17. Y x F3(F2) x D 15 3.768 0.251 0.095 15 264*
18. Y x L X F2 14 7.629 0.544 0.440 15 1.23
19 Y x L x F3(F2) 15 6.602 0.440 0.205 236 2.14**
20. Y x L x D 1 1.003 1.003 0.185 14 541*
21. L x F2 x D 14 1.757 0.125 0.250 13 0.50
22. L x -F3(F2)xD 15 2.396 0.159 0.095 15 1.68
23. Y x L x F2 x D 14 2.598 0.185 0.095 15 1.95
24. YxLxF3(Fz)xD 15 1.424 0.095 0.205 236 0.46
25. Error 236 48.549 0.205

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 13.5%
Mean = 3.35

 

Appendix C.2. Analysis of variance table
for Maturity.

127

of the Nested model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# Source df SS MS MSe dfe F
1. Year(Y) 1 4838.7 4838.7 2667.9 1 1.18
2. Location(L) 1 2511.6 2511.6 2657.5 0 0.94
3. Y xL 1 2660.2 2660.2 45.8 10 58.06**
4. Rep(YL) 4 104.1 26.0 3.1 236 8.52**
5. F2 14 3395.7 242.5 90.0 15 2.69*
6. F3(F2) 15 1115.9 74.4 12.2 5 6.07*
7. Pub.Density(D) 1 409.6 409.6 117.0 0 3.50
_§_._ I“2 x D 14 460.2 32.8 44.8 13 0.73
9. F3(F2) x D 15 681.0 45.4 5.9 9 7.73**
10. L x F2 14 282.3 20.2 21.4 8 0.94
11. L x F3(F2) 15 145.1 9.6 11.1 15 0.87
12. Lx D 1 0.0 0.0 4.5 1 0.00
13. Y x F2 14 427.5 30.5 13.6 15 2.24
14. Y x F3(F2) 15 204.3 13.6 22.8 14 0.59
15. YxD 1 121.7 121.7 4.4 1 27.41
16. Y x F2 x D 14 58.3 4.1 4.4 6 0.94
17. Y xF3(F2) xD 15 72.7 4.8 3.5 15 1.38
18. YxLXF2 14 319.6 22.8 11.1 15 2.06
19 Y x L x F3(F2) 15 165.8 11.1 3.1 236 3.61**
20. YxLxD l 3.4 3.4 3.1 14 1.08
21. LxF2xD 14 59.9 4.3 4.1 5 1.03
22. L x F3(F2)xD 15 67.7 4.5 3.5 15 1.29
23. Y x L x F2 x D 14 43.3 3.1 3.5 15 0.88
24. YxLxF3(F2)xD 15 52.4 3.5 3.1 236 1.14
25. Error 236 720.8 3.1

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 5.8 %
Mean = 30.3

 

128

Appendix C.3.' Analysis of variance table of the Nested model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

for Height.
# Source df SS MS MSe dfe F
1. Year(Y) 1 238.0 238.0 1418.6 1 0.16
2. Location(L) 1 8151.0 8151.0 1298.7 1 6.27
3. Y xL 1 1267.5 1267.5 2391.6 3 0.53
4. Rep(YL) 4 9642.1 2410.5 68.6 236 35.09**
5. F2 14 26663.9 1904.5 743.1 15 2.56*
6. F3(Fz) 15 9937.1 662.4 151.6 7 437*
7. Pub.Density(D) 1 18924.2 18924.2 277.1 0 68.27
8. F2 x D 14 4737.2 338.3 514.6 14 0.65
9. F3(F2) x D 15 7367.3 491.1 96.8 9 5.07**
10. L x F2 14 1134.9 81.1 25.1 0 3.23
11. L x F3(F2) 15 1061.5 70.7 95.4 15 0.74
12. L x D 1 168.6 168.6 474.9 1 0.35
13. Y x F2 14 2814.0 201.0 176.3 15 1.14
14. Y x F3(F2) 15 2645.5 176.3 49.8 14 3.53
15. Y x D 1 18.0 18.0 520.3 1 0.03
16. Y x F2 x D 14 1380.8 98.6 79.5 7 1.23
17. Y x F3(Fz) x D 15 1454.4 96.6 49.0 15 1.97
18. Y x L X F2 14 698.1 49.8 95.5 15 0.52
19 Y x L x F3(Fz) 15 1433.1 95.5 68.6 236 1.39
20. Y x L x D 1 453.3 453.3 31.6 14 14.32**
21. L x F2 x D 14 746.2 53.3 31.5 2 1.69
22. L x F3(F2)xD 15 733.4 48.8 49.0 15 0.99
23. Y x L x F2 x D 14 443.0 31.6 49.0 15 0.64
24. YxLxF3(Fz)xD 15 735.3 49.0 68.6 236 0.71
25. Error 236 16211.1 68.6

 

 

 

 

 

 

 

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 8.1%
Mean = 103.1

 

129

Appendix C.4. Analysis of variance table of the Nested model

for Lodging.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# Source df SS MS MSe dfe F
1. Year(Y) 1 48.13 48.13 0.017 0 2720.4*
2. Location(L) 1 9.08 9.08 0.725 3 1250*
3. Y xL 1 0.00 0.00 8.021 4 0.00
4. Rep(Y,L) 4 31.05 7.76 0.328 236 23.60**
5. F2 14 99.46 7.10 3.884 17 1.82
6. F3(F2) 15 46.58 3.11 0.551 5 5.63*
7. Pub.Density(D) 1 28.78 28.78 1.851 1 15.54
8. F2 x D 14 13.37 0.95 1.196 10 0.79
9. F3(F2) x D 15 16.35 1.09 0.351 3 3.11
10 L x F2 14 18.39 1.31 0.554 5 2.37
11. L x F3(F2) 15 7.48 0.50 0.532 15 0.937
12. L x D 1 0.79 0.79 0.181 2 4.40
13. Y x F2 14 8.48 0.61 0.585 15 1.03
14. Y x F3(F2) 15 8.78 0.59 0.587 14 0.99
15. Y x D 1 1.45 1.45 0.072 0 19.96
16. Y x F2 x D 14 3.85 0.27 0.006 0 40.59
17. Y x F3(Fz) x D 15 2.87 0.19 0.386 15 0.49
18. Y x L X F2 14 8.23 0.59 0.532 15 1.10
19 Y x L x F3(Fz) 15 7.98 0.53 0.328 236 1.61*
20 Y x L x D 1 0.00 0.00 0.201 14 0.00
21. L x F2 x D 14 5.36 0.38 0.361 3 1.06
22. L x F3(F2)xD 15 8.18 0.55 0.386 15 1.41
23. Y x L x F2 x D 14 2.83 0.20 0.386 15 0.523
24. YxLxF3(Fz)xD 15 5.79 0.37 0.328 236 1.17
25. Error 236 77.62 0.33

 

*,** Significant at p>0.05 and P>0.01, repectively.

C.V. =

26.2%

Mean = 2.22

 

130

Appendix C.5. Analysis of variance table of the Nested model

for Seed size.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sburce df SS MS MSe dfe F
1. Year(Y) 1 1327.5 1327.5 58.92 0 22.53
2. Location(L) 1 70.3 70.3 59.23 0 1.18
3. Y xL 1 63.8 63.8 26.78 6 2.38
4. Rep(Y,L) 4 84.2 21.1 0.60 236 34.93**
5. F2 14 221.7 15.8 5.63 3 2.81
6. F3(F2) 15 131.8 8.7 0.02 0 413.04*
7. Pub.Density(D) 1 4.2 4.2 9.94 1 0.42
8. szD 14 100.2 7.1 4.71 7 1.52
9. F3(F2) x D 15 93.5 6.2 1.48 11 420*
10. L x F2 14 24.5 1.7 5.51 9 0.38
11. L x F3(F2) 15 19.6 1.3 2.12 15 0.61
12. L x D 1 5.2 6.9 1.3 4 5.02
13. Y x F2 14 20.1 1.4 0.82 15 1.73
14. Y x F3(F2) 15 12.4 0.8 6.32 14 0.13
15. YxD 1 0.6 2.2 1.16 3 1.92
16. Y x F2 x D 14 10.3 0.7 1.88 12 0.39
17. Y x F3(Fz) x D 15 12.1 0.8 0.66 15 1.21
18. Y x L X F2 14 88.5 6.3 2.12 15 2.98*
19 Y x L x F3(F2) 15 31.8 2.1 0.60 236 3.51**
20. YxLxD 1 0.42 0.42 1.73 14 0.24
21. L x F2 x D 14 13.4 0.95 2.41 15 0.39
22. L x F3(Fz)xD 15 20.1 1.33 0.66 15 2.02
23. Y x L x F2 x D 14 24.3 1.73 0.66 15 2.62*
24. YxLxF3(Fz)xD 15 9.92 0.66 0.60 236 1.09
25. Error 236 142.2 0.6

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 4.1%
Mean = 19.24

 

131

Appendix C.6. Analysis of variance table of the Nested model

for Protein.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# Source df SS MS MSe dfe F
1. Year(Y) 1 5447.26 5447.26 6.58 0 827.85*
2. Location(L) 1 19.76 19.76 7.57 1 2.61
3. Y xL 1 6.96 6.96 1.12 14 6.18*
4. Rep(Y,L) 4 0.05 0.012 0.006 236 2.00**
5. F2 14 132.90 9.49 1.52 3 6.24*
6. F3(F2) 15 27.07 1.80 1.63 29 1.10
7. Pub.Density(D) 1 3.48 3.48 2.15 1 1.61
8. F2 x D 14 20.53 1.47 0.35 0 4.20
9. F3(F2) x D 15 13.63 0.91 1.13 20 0.80
10. L x F2 14 24.22 1.73 1.99 27 0.86
11. L x F3(F2) 15 13.34 0.89 0.02 15 4450*
12. L x D 1 0.00 0 1.86 0 0.00
13. Y x F2 14 10.36 0.74 0.76 15 0.97
14. Y x F3(F2) 15 11.47 0.76 1.12 14 0.67
15. Y x D 1 0.00 0.00 2.09 1 0.00
16. Y x F2 x D 14 10.41 0.74 1.09 19 0.67
17. Y x F3(Fz) x D 15 9.37 0.62 0.21 15 2.95*
18. Y x L X F2 14 15.68 1.12 0.02 15 56.00**
19 Y x L x F3(Fz) 15 0.34 0.02 0.006 236 3.33**
20. Y x L x D 1 2.03 2.03 0.68 14 2.98
21. LxF2xD 14 7.16 0.51 1.19 20 0.42
22. L x F3(F2)xD 15 10.73 0.72 0.21 15 342*
23. Y x L x F; x D 14 9.51 0.68 0.21 15 323*
24. YxLxF3(F2)xD 15 3.12 0.21 0.006 236 35.00**
25. Error 236 1.41 0.006

 

 

 

 

 

 

 

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 0.2 %
Mean = 39.2

 

132

Appendix C.7. Analysis of variance table of the Nested model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

for Oil.

# Sorn'ce df SS MS MSe dfe F

1. Year(Y) 1 126.58 126.58 1.04 5 121.71*

2. Location(L) 1 0.33 0.33 0.54 1 0.61

3. Y xL 1 0.35 0.35 0.35 13 0.98

4. Rep(YL) 4 0.011 0.003 0.006 236 0.50

5. F2 14 30.19 2.15 1.74 16 1.23

6. F3(Fz) 15 13.88 0.93 0.43 10 2.16

7. Pub.Density(D) 1 2.72 2.72 3.83 0 0.71

__8_._ F2 x D 14 4.04 0.29 0.31 3 0.93

9. F3(Fz) x D 15 5.72 0.38 0.21 6 1.81
10. L x F2 14 7.74 0.55 0.36 8 1.52
11. L x F3(F2) 15 2.93 0.20 0.20 15 1.00
12. L x D 1 0.153 0.40 0.23 6 1.73
13. Y x F2 14 14.72 1.05 0.43 15 2.44
14. Y x F3(F2) 15 6.38 0.43 0.36 14 1.19
15. Y x D 1 3.62 3.88 0.22 6 17.63**
16. Y x F2 x D 14 2.66 0.19 0.26 7 0.73
17. Y x F3(Fz) x D 15 2.74 0.18 0.17 15 1.06
18. Y x L X F2 14 4.97 0.36 0.20 15 1.80
19 Y x L x F3(Fz) 15 2.96 0.20 0.006 236 33.33**
20. Y x L x D l 0.03 0.03 0.25 14 0.12
21. L x F2 x D 14 2.68 0.20 0.28 8 0.71
22. L x F3(Fz)xD 15 2.91 0.20 0.17 15 1.17
23. Y x L x F2 x D 14 3.55 0.25 0.17 15 1.47
24. YxLxF3(F2)xD 15 2.607 0.17 0.006 236 28.33**
25. Error 236 1.474 0.006

 

*,** Significant at p>0.05 and P>0.01, repectively.
C.V. = 0.4 %
Mean = 19.2

 

Appendix D.l. Estimates of correlation coefficients and their

133

respective significance among variables by Year.

 

 

 

 

 

 

 

 

Variable Yield maturity Height Lodging Seed size Protein Oil
Yield

'89 _ 026“ 0.04 -0.02 0.37" 0.48“ -0.27**

'90 _ 0.51“ 017“ -0.08 0.43" 0.21 *"' 0.03

AVE ___ 0.27“ -0.06 0.01 0.38" 0.24“I 0.03

Maturity

'89 __ 0.45" 0.36" 030" -0.47** 0.24"

'90 _ 0.36“ 0.37" 0.30“ -0.04 016“

AVE __ 0.28“ 0.45" 0.44“I 0.45“ 0.32“
Height

'89 0.61" -0.02 023*" -0.02

'90 0.59“ 020*“ 017“ 026"

AVE __ 0.55“I -0.11‘ -0.09" -0.l4”

Lodging

'89 _ 0.05 -0.12 0.11

'90 _ 0.03 -0.09 -0.19""'

AVE 0.26" 0.28" 0.17“"I
Seed size

'89 __ 0.29“ -0.09

'90 _ 0.31“ -0.02

AVE ___ 0.75” 0.50“I
Protein

'39 _ 063**

'90 _ 060“

AVE __ 0.59“
Oil

'89 _-

‘90 _

AVE _

 

*,""" Significant at p<0.05 and p<0.01, respectively.

Appendix D.2. Estimates of correlation coefficients and their

134

respective significance among variables by Location.

 

 

 

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil
Yield

ING -0.15"' -0.12 -0.03 0.31“ 0.12 -0.02

LEN _ 0.24“ 0.26” 0.20“ 0.38" 0.36"”I 0.03

AVE _____ 0.27“ -0.06 0.01 0.38“I 0.24“ 0.03
Maturity

ING 0.62“ 0.48M 0.01 0.11 -0.02

LEN 0.30“ 0.60" 0.64" 0.76“ 0.53“

AVE ___ 0.28“I 0.45“I 0.44” 0.45“ 0.32"
Height

ING 0.39“ -0.27"'" -0.20“ 028"

LEN 0.66" 0.07 0.03 -0.03

AVE __ 0.55"I -0.ll“ -0.09" -0.14“
Lodging

ING _ 0.35"”' 0.32“ 0.16“

LEN 0.24" 0.27“ 0.19“

AVE 0.26” 0.28"I 0.17”
Seed size

ING 0.71" 0.44'MI

LEN __ 0.81""' 0.54“

AVE __ 0.75“I 0.50“
Protein

ING _ 0.67"

LEN _ 0.54"

AVE __ 0.59“
Oil

ING ING: Ingham _

LEN LEN: Lenawee _

AVE ‘, ’”' Significant at p < 0.05 and p< 0.01, respectively. _

 

Appendix D.3. Estimates of correlation coefficients and their respective

135

significance among variables by Year x Location.

 

 

 

 

 

 

 

 

Variable Yield Maturity Height Lodging Seed size Protein Oil

Yield

'89 ING __ 039“ -0.15 -0.05 0.49" 0.50""I -0.29"‘I
[EN _ -0.l7‘I 0.45” 0.13 0.30“ 0.27“ -0.29“

'90 ING _ -0.02 -0.06 -0.64 0.09 -0.09 0 .11
[EN _ 0.16 -0.07 0.09 0.22“I 0.32“ -0.65“
AVE __ 0.27” -0.06 0.01 0.38” 0.24” 0.03

Maturity

'89 ING _ 0.69“l 0.13 -0.61” -0.53“ 0.02
[EN _ 0.33“ 0.54“ 0.04 -0.46‘"' 0.41“

'90 ING _ 0.69“ 0.57“ 0.12 -0.l7“ -0.3l “
[EN _ 0.59“l 0.70“ -0.14 -0.15 -0.33"'"
AVE ____ 0.28” 0.45“ 0.44“ 0.45“ 0.32”

Height

'89 ING _ 0.37” -0.47"‘I -0.31"‘I -0.13
LEN _ 0.73“ 0.38" 0.03 0.05

'90 ING _ 0.55" 0.07 -0.07 -0.31“'
[EN _ 0.63“ -0.31"‘I -0.22""' -0.22

AVE ___ 0.55“ -0.11’ -0.09‘ 41.14”

Lodging

'89 ING _ -0.02 0.05 -0.08
LEN _ 0.13 -0.17 0.24""I

'90 ING _ 0.41” 0.05 -0.15
LEN _ -0.19“ -0.19‘ -0.23“
AVE __ 0.26” 0.28“ 017*-

Seed size

'89 ING _ 0.47" -0.15
LEN _ 0.09 003

‘90 ING _ 0.27” -0.09
IEN _ 0.3227“I -0.05
AVE __ 075*- o.so**

Protein

'89 ING _ ~05?“
[EN _ -0.78”

'90 ING _ -0.55'"'
IEN __ 0.65"
AVE __ 059* *

Oil

'89 mo ING: rnghun __
[EN IEN: Lenawee __

’90 ING _
IEN _
AVE

 

‘, " Significant at p < 0.05 and p< 0.01, respectively

"111111111111111111“