WI HI] I lWWIUH‘TIWIWFWIW L

g 3 1293 00658 7178

O
y' . 0":
o‘ '
. /3
..~\
«:14;
.«‘ 1 I. l.’ -
1
1

ABSTRACT
EVALUATION OF RESPONSE TO SELECTION AND MATING
SYSTEMS NITH DIFFERENT LEVELS OF HERITABILITY.
SELECTION INTENSITY AND PROGENY TESTING THROUGH
THE USE OF SIMULATION

Daryl Ralph Strohbehn

The data used in this study were collected from the simulation of
approximately 295,000 animals. A biological model was used to simulate
the performance of animals based on genic effects from the parents,
Mendelian sampling and environmental effects. The selections and
mating of all animals were accomplished by using different FORTRAN IV
computer programs.

Random mating (RM) and phenotypic positive assortative mating (PAM)
were the two mating systems compared, along with two bull testing
procedures, within the purebred herd and with a commercial test herd.
Other independent variables included in the study were three levels
of yearling weight heritability (YWH), 20, 40 and 60%, three different
bull combinations, 0 progeny tested and 6 non-progeny tested, 2 progeny
tested and 4 non-progeny tested, and 3 progeny tested and 3 non-progeny
tested, three levels of percent heifer calves saved (PHCS) 20, 50 and
80%, and three selection criteria, estimated breeding value of yearling
weight (EBV), yearling weight (YN), and average daily gain (ADG). When
selection was practiced on vw or A00, the bull and heifer calves were
selected on their phenotypic expression, while the older, producing

bulls and cows were selected totally on progeny averages.

Traits investigated in the study were phenotypic and genic
means and variances for weaning weight (NH), A06 and Y” for calves,
sires and dams. Also summarized were EBV averages for calves, sires
and dams, plus progeny means for parents and average number of off-
spring per parent. Genetic and phenotypic correlations between
mated parents for A00 and YN, plus the correlation between EBV's
of mated parents were analyzed.

A two part analysis of variance was done, with one procedure
analyzing the two mating systems and the four factorialized sources
of variance, while the second procedure analyzed the two sire testing
procedures and the four factorialized sources of variance.

PAM and RM were not significantly different.in the amount of
improvement they caused in YH over ten calf crops. PAM caused an
insignificant increase in phenotypic and_genic variance for YH. The
phenotypic correlation for YN between mated individuals for PAM was
.88, while -.06 for RM. Even though the difference in YH genetic
correlation between mates for the two mating systems was significant
(P<.0l), it was much smaller, .16 for PAM versus -.06 for RM.

PAM made superior genetic and phenotypic gains in ADG over RM.
PAM populations at calf crop 10 averaged .04 lbs. per day higher than
RM populations in ADG. Furthermore, PAM caused significantly greater
genic and phenotypic variance at calf crops 4 and 10. The average
phenotypic and genetic correlations between mates for the three calf
crops for ADG selection in PAM were .90 and .30, respectively, while

in RM populations the same respective figures were .02 and .01.

Sire testing within a commercial test herd, as done in this
investigation, was inferior to purebred herd sire testing for genetic
improvement of all performance traits monitored. The populations
utilizing the commercial test of sires were 98.4 and 97.3% as heavy
as the other sire testing procedure at yearling time in calf crop
4 and 10, respectively. In calf crop l0, the populations utilizing
commercial testing of sires were 97.5 and 97.0% as heavy for NH and
A06, respectively, when compared to purebred herd testing for sires.

The means for the three YNH levels, for all performance traits
monitored, appeared to be significantly different from one another.
NH, A06 and Y" reSponded in a linear fashion to increases in YNH,
while genic maternal ability was not affected by YNH changes.

The different bull combinations utilized in this study caused
no significant change in the amount of genetic improvement made in
YN over time. However, data were presented to show that the combina-
tion of 2 progeny tested and 4 non-progeny tested bulls gave the
highest averages for the three calf cr0ps and three traits analyzed,
with the exception of A06 in calf crop 7 of the mating system analysis.

Increased heifer saving rates had a positive effect on YH
improvement. Fifty and 80% levels were significantly (P<.Ol) greater
in YH improvement than the 20% retention rate. Direct genic value
for NH was the highest for the herds incorporating 50 and 80% saving
rates. The genic maternal ability increased as PHCS increased up to
50%, however, a leveling off occurred in the mating system analysis
and a nonsignificant depression in the sire testing analysis occurred

when PHCS changed from 50 to 80%.

The two higher PHCS levels did not differ significantly from
one another in ADG improvement, however, both were significantly
superior to 20% PHCS in genic and phenotypic averages. The 20% level
of PHCS had 19% more phenotypic variance and 34% more genic variance
at calf crap l0 of the mating system analysis than 50 and 80% PHCS.

The unadjusted NW was higher for the 20% PHCS level through
calf crop 7. However, by calf crop lO, 50 and 80% PHCS levels were
superior due to their advantage in genic maternal ability and direct
genic value for NW. The unadjusted YN averages were significantly
higher for 50 and 80% PHCS levels than for 20% by calf crop 7
because of superiority in genetic gain for ADG.

Generation interval was affected to a major degree by level of
PHCS. At calf crop 10 of the mating system analysis, the calculated
cow turnover rates were 7.6, 5.l and 4.l years, respectively, for 20,
50 and 80% PHCS.

Selection by YH or EBV caused the most improvement in YN. Through
calf crop 7, EBV selection consistently gave greater improvement in Y",
but by calf crop l0, Yw selection produced heavier calf YH. However,
EBV and YH selection did not differ significantly from one another.

A possible, serious bias in older female EBV was found. The EBV
of a selected female is biased upward in herds that are rapidly improv-
ing in performance. This occurs because the females' individual
deviation from herd mean was not readjusted, plus the fact that the
records for calves produced in the early part of a cows' productive

life are included in the EBV calculation as deviations from herd means

Ethat are below the herd genetic mean at the time of older cow
selection.

YR selection caused the greatest amount of correlated improve-
ment in direct genic value for WW and genic maternal ability.

ADG selection had significantly (P<.Ol) greater improvement
in ADG when compared to EBV and Yw selection, which were not
significantly different from one another.

Many interactions between the independent variables occurred,
however, five of the seven discussed in the thesis involved YNH.

One of these was the interaction of mating system and YNH for ADG.
ADG was equal in RM and PAM when YWH was 20 and 40%, however, with
60% YHH, ADG averaged .07 lbs. per day higher in PAM than with RM.

Sire testing interacted with YNH in calf crop 10 for A06 and
vw. A greater increase in both A06 and Y" occurred as YHH increased
for the case of within purebred herd sire testing than for sire'
testing in a commercial test herd. Sire testing also interacted with
PHCS for YH and unadjusted vw in calf crop l0. It resulted because
sire testing within the purebred herd responded to increases in PHCS
more than sire testing within a commercial test herd.

YWH interacted significantly with selection criteria in NH, ADG.
and unadjusted NH in both analyses and in all three calf crops
analyzed. Also it occurred for genic maternal ability in calf crops
7 and 10 of the mating system analysis. ADG selected populations did

not respond as much to YNH increases as Y“ and EBV selected populations

for the above traits. However, in the mating system analysis, ADG
selection gave more improvement in ADG than either EBV or vw
selection as YHH increased.

The interaction of YWH and PHCS was significant for YN and A06
in all three calf crops of the mating system analysis. Improvement
increased linearly across PHCS levels with 20 and 40% YNH, however,
with the 60% level of YWH, a leveling off occurred in A06 and YN
when PHCS changed from 50 to 80%. Significance of the interaction,
with the same action as described in the previous sentence, occurred
in genic value for NH in calf crops 7 and 10, but only in calf crop
7 for NH phenotype.

The final independent variable YWH interacted with was bull
combination. At low YHH in the mating system analysis more improve-
ment in ADG was made when more progeny tested bulls were incorporated
into the breeding system. However, at the highest YHH level the use
of fewer progeny tested bulls enhanced the rate of improvement.

The last interaction discussed in the thesis was between
selection criteria and PHCS. It was significant for A00 in the mating
system data. EBV and YN selected populations had an average improve-
ment in ADG of .08 lbs. per day when PHCS increased from 20 to 80%,
while ADG selected populations improved .20 lbs. per day as PHCS
went from 20 to 80%.

EVALUATION OF RESPONSE TO SELECTION AND MATING

SYSTEMS NITH DIFFERENT LEVELS OF HERITABILITY,

SELECTION INTENSITY AND PROGENY TESTING THROUGH
THE USE OF SIMULATION

By
Daryl Ralph Strohbehn

A THESIS

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

DOCTOR OF PHILOSOPHY

Department of Animal Husbandry

1974

ACKNOWLEDGEMENTS

The author wishes to express his sincere thankfulness for
the many dedicated, unselfish and exuberant hours of guidance
and labor his major professor, Dr. N. T. Magee, put forth in both
the author's personal growth and academic training. Indeed
without his help and encouragement the amount of toil and hardship
would have been magnified.

The author desires to acknowledge and thank Dr. R. H. Nelson
and the Animal Husbandry Department for the financial assistance
given to him throughout his graduate tenure in the way of a graduate
assistantship. Hearty thanks also go out to the consultants at
the Michigan State Computer Laboratory fer their help and under-
standing in achieving the simulation work reported on in this
document.

Last, but not least, the author solicits his greatest amount
of gratitude to his parents, Mr. and Mrs. Ralph R. Strohbehn, fer
their never-ending advice and encouragement throughout life and

especially while in graduate school.

ii

TABLE OF CONTENTS

 

Page
LIST OF TABLES ........................ vi
LIST OF FIGURES ....................... viii
INTRODUCTION ......................... 1
REVIEW OF LITERATURE ..................... 5
Methods of Selection and Their Impact
on Improvement .................... 5
Results of Selection Experiments ............ 8
Selection and Improvement in Simulation
Studies ....................... l5
Mating Systems and Their Influence on
Populations ..................... 23
Estimating Breeding Value ............... 3O
Estimates of Genetic Parameters ............ 42
Heritabilities of Traits Studied ......... 42
Phenotypic, Genic and Environmental
Variances of the Traits Studied ......... 45
Correlations Between Traits Studied ........ 47
METHODS AND PROCEDURES OF SIMULATION ............. 49
Introduction ...................... 49
Simulation of Herd Populations ............. 53
Mechanics of Selection Programs ............ 61
Mechanics of Mating Programs .............. 63
PARAMETER VALUES USED .................... 65
RESULTS AND DISCUSSION .................... 71
The Effects of Main Sources of Variance ........ 7l
Mating System ................... 71
Effect on Yearling Height .......... 7l
Effect on Weaning Weight ........... 75

iii

TABLE OF CONTENTS--continued Page

RESULTS AND DISCUSSION--continued

Effect on Average Daily Gain ......... 76
Effect on Estimated Breeding Value ...... 77
Sire Testing Procedures .............. 80
Effect on Yearling Weight .......... 80
Effect on Weaning Weight and Average
Daily Gain ................. 80
Yearling Weight Heritability ........... 81
Effect on Performance Traits ......... 81
Effect on Estimated Breeding Value and
Number of Offspring Per Parent ....... 82
Bull Combination ................. 83
Effect on Performance Traits ......... 83
Effect on Estimated Breeding Value ...... 84
Percent Heifer Calves Saved ............ 85
Effect on Yearling Weight .......... 85
Effect on Weaning Weight ........... 86
Effect on Average Daily Gain ......... 86
Effect on Unadjusted Weaning Weight
and Yearling Weight ............ 89
Effect on Generation Interval ........ 90
Selection Criteria ................ 91
Effect on Yearling Weight .......... 91
Effect on Weaning Weight ....... '. . . . 92
Effect on Average Daily Gain ......... 94
Effect on Unadjusted Weaning Weight
and Yearling Weight ............ 94
Effect on Estimated Breeding Value ...... 95
General ................... 96
Interactions Between Main Sources of Variance ..... 96
Mating System and Yearling Weight
Heritability .................. 96
Sire Testing and Yearling Weight
Heritability .................. 99
Sire Testing and Percent Heifer
Calves Saved .................. 99
Yearling Weight Heritability and
Selection Criteria ............... lOl
Yearling Weight Heritability and
Percent Heifer Calves Saved ........... 103
Yearling Weight Heritability and
Bull Combination ................ 105
Selection Criteria and Percent
Heifer Calves Saved ............... 105
General ...................... 106

iv

TABLE OF CONTENTS--continued

APPENDIX A ..........................
BIBLIOGRAPHY .........................

Page

107
112
119
120

LIST OF TABLES

TABLE Page

d
.

Abbreviations, Values and Definitions of Factors
Used in Forming Genotypes and Phenotypes for
Simulation of Herd Populations Section ........ 55

2. Components Required to Prescribed Levels of
Yearling Weight Heritability ............. 69

3. Analysis of Variance and List of Means for Main
Sources of Variance in Mating System Data for
WW, ADG and YW .................... 73

4. Analysis of Variance and List of Means for Main
Sources of Variance in Sire Testing Data for
WW, ADG and YW .................... 78

5. Analysis of Variance and List of Means for Main
Sources of Variance for EBV in Mating System
and Sire Testing Data ................ 79

6. Calf Average for Genic Maternal Value in Mating
System and Sire Testing Data .......... .. . . 87

7. Bull Calf Average for Unadjusted Weaning Weight
in Mating System and Sire Testing Data ........ 87

8. Bull Calf Average for Unadjusted Yearling Weight
in Mating System and Sire Testing Data ........ 88

9. Means and Significance Values of EBV and YW
Breeding Values for Sires and Dams in Mating
System and Sire Testing Data ............. 88

10. ADG Means for Mating System and YW Heritability
Subclasses ...................... 98

ll. EBV Means for Mating System and YW Heritability
Subclasses ...................... 98

12. ADG and YW Means in Calf Crop 10 for Sire Testing
and YW Heritability Subclasses ............ 100

vi

LIST OF TABLES--continued Page

13. YW and Unadjusted YW Means in Calf Crop 10 for
Sire Testing and Percent Heifer Calves Saved _
Subclasses ...................... 100

14. WW and MA Means in Calf Crop 10 for YW Heritability
and Selection Criteria for Mating System
Data ......................... 102

15. ADG Means for the Interaction of YW Heritability
With Selection Criteria for Mating System
Data ......................... 102

16. ADG and YW Means for YW Heritability and Percent
Heifer Calves Saved in Mating System
Data ......................... 104

17. ADG Means for YW Heritability and Bull Combina-
tions in Mating System Data ............. 104

vii

LIST OF ILLUSTRATIONS
FIGURE Page
1. Flow Chart of Programs Utilized to Simulate

Beef Herd Populations ................ 52

2. Diagram of the Factors Influencing Performance
Traits in Simulated Animals ............. 57

viii

INTRODUCTION

The quest of mankind for years has been to breed our domesticated
animals for efficiency of meat and milk production. Because goals
change with time, beef production has been altered immensely which
retroactively causes changes in breeding schemes and the amount of
emphasis placed on traits which vary in heritability and accuracy
and ease of measurement.

In today's beef industry it is generally an accepted fact that
purebred breeders must raise breeding animals that will satisfy
the production segments of the industry. The animal must be an
efficient converter of feedstuffs into pounds of live weight at
all stages of development. Breeders further realize that to become
divorced from consumers' demands of lean, high quality beef is
archaic and one which could destroy the future of the beef industry.
Due to these facts, much effort is being put forth by breeders to
learn more about new, scientific approaches to selection and mating
problems. However, many of the more scientific approaches to
selection and mating have not been utilized in improvement programs
long enough to give answers, thus scientists in many cases are
making educated guesses at what problems and ramifications might

result from these approaches.

In order to restore confidence to our extension activities in
the field of animal breeding, data must be made available which will
either back up or disspell certain theories and methodology being
incorporated in breeding schemes. But to overcome the obstacle of
data collection is not easy, indeed it is expensive and very time
consuming. In fact, it can be so time consuming, that many breeders
may utilize a theoretically sound scheme which ends up after years
of data collection to have less merit than other simpler or more
complex approaches.

A partial solution to the problem of data collection is the
use of a production model which employes the use of a high-speed
computer, a well-developed algorithm and combinations of many
factors controllable by a breeder. With the use of this type of
system one can seek answers on what possible combination will allow
optimum genetic improvement in performance traits. Indeed to
acquire informative data requires the use of proper methodology and
establishment of correct parameters. Without the first, the proper
parameters will be to no avail, and likewise, without proper
parameters, the correct methodology will give no help in finding
solutions to dilemmas breeders face.

The main objectives of this investigation will be to establish
trends with simulated populations which may need to be further
researched at greater length. Theobjectives as seen fer this
investigation are:

1. Compare how random mating and phenotypic positive

assortative mating systems affect beef populations.‘

Phenotypic and genetic trends will be monitored
along with changes which may occur in phenotypic

and genic variance at three heritability levels.
Investigate what differences occur in improvement
of performance traits when two sire testing
procedures are used.

Compare different bull turnover rates in the amount
of improvement they cause.

Study the potential involved in saving different
fractions of the heifer calves produced. Both age
of dam adjusted and unadjusted weight trends will

be evaluated to find in what range of heifer replace-
ment rates lies maximum improvement and minimum
weight loss due to age of dam adjustment.

Compare the results obtained when selection criteria
is varied. An investigation into the advantages and
disadvantages of selection by estimated breeding
value for yearling weight will be done, along with
comparing it to selection by yearling weight and
average daily gain. The effects these have on
phenotypic and genic means and variances will be
analyzed.

Three levels of heritability will be used. This
should help to extrapolate the results of other

traits.

7.

The last objective of the investigation is to study
what interactions occur in the above independent
variables and whether they cause any serious
detrimental effects on improvement of performance

traits.

REVIEW OF LITERATURE

Without a doubt there are many aSpects involved with genetic
improvement and to survey them in totality would be exhausting and
quite unnecessary in regards to the task of this work. Indeed a
review of many t0pics is needed for this investigation, but light
coverage is given to some areas because documentation of research
is slight and often sketchy in those areas. Therefore, emphasis
will be placed on the important aspects to this research with a
degree of completeness to the following three t0pics: Methods of
Selection, Mating Systems, and Estimates of Genetic Parameters.

"Methods of_$election and Their Impact
gg_Improvement

 

 

Selection has been given many definitions, but one which most
animal breeders can live with was given by Lush (1948, p. 171). He
defined it as differential reproductive rates. His more precise defini-
tion is "... number of offspring which reach breeding age for each
individual which reached breeding age in the preceding generation."
Falconer (1960, p. 26) similarly describes selection as "the pr0por-
tionate contributions of off5pring to the next generation which is also
called the fitness of the individual, or sometimes the adaptive value,
or selective value.“ Many complications arise in quantitatively

describing selection. Differences in viability and fertility can

cause a diSproportionate effect on selection and bring about discrete
changes in intensity of selection which usually amount to a slow down
in genetic improvement. However, the important thing to observe is
that, even though natural selection has an active role, man can cause
differential reproductive rates and bring about genetic changes. Of
course, this genetic change is produced by changing gene frequency
which for performance traits appears to be a small alteration at each
of a large number of loci. Gene frequency changes have been theorized
and published for cases of different reproductive rates, dominant gene
action, overdominance, etc., but because of its basic unimportance in
this work these will not be reviewed.

Truncation selection is the most extreme type possible once a
culling level is established. When a normal distribution is truncated
the mean of the truncated p0pulation is z/b-o from the mean of the
whole population. The b is the fraction of the p0pulation selected
and z is the height of the ordinate at ko. The magnitude of average
selected parent superiority (or selection differential magnitude) is
dependent upon the pr0portion saved and the phenotypic standard
deviation. With smaller papulation variances less genetic reach can
be practiced resulting in decreased rates of improvement. Because
of truncation selection's severity most breeders use modified versions
which allow them to bring up animals not meeting certain requirements,
but excelling in other traits. If truncation selection is adapted by
a breeder it usually is used when an index is formed utilizing data on

economically important traits.

The z/b figures given by Lush (1945) are exact only for truly
normal curves. Since departures from normality are common these
figures will not be exact in many cases. When skewed toward low
merit small amounts of truncation will have more effect and heavy
truncation will have less effect than if the distribution were truly
normal. If the curve is skewed in the high merit direction the
Opposite will fit. Lush (1948) further explains that if the
distribution curve is flatter, very small and very large amounts of
truncation will have greater effects than in normal distributions,
while moderate truncation will give little difference. However, if
more individuals are concentrated about the average and few are in
the extremes, extremely light and heavy culling will cause minimal
change while intermediate levels will bring about greater changes
than they would when the p0pu1ation is normally distributed.

Random environmental variations along with dominance and
epistasis affect the rate at which selection changes gene frequency.
The former acts to Spread genic classes out over several phenotypic
classes, thus creating situations which bring about selection errors.
Dominance deviations cause animals to breed different than their
phenotypes indicate. Dominance deviations are like random environ-
mental variations, in that they cannot be transmitted from parent to
offSpring. Epistatic variations are transmitted to a varying degree.
Some part of whatever epistatically caused differences in the parents

will show up temporarily in the offSpring.

Accomplishment of small selection differentials or low
heritability, or the combination of both, can act as road blocks
to genetic improvement. Lush (1948) explains that selection
differential can be small due to p0pu1ation uniformity, low
reproductive rates and poor selection practices. Heritability
may be small because of small genic variance, large environmental
effects, dominance deviations of large magnitude, large epistatic
variance, or the variance from non-linear interactions between
environment and heredity may be large. Heritability cannot usually
be changed by the breeder, thus the emphasis of this work will be
to investigate different selection routines with Specific levels of

heritability.

Results gf_Selection Experiments

 

Various results from selection have been observed in beef cattle.
However, it appears variation due to inbreeding is incorporated in
many of the studies. Brinks (1965) reported increases of 5.8, 6.3
and 2.37 lbs. per year for final test weight, l96-day gain and weaning
weight, respectively. Bulls were sequentially culled on the basis of
weaning weight and score, l96-day gain, and in most instances a
progeny test while females were sequentially culled on 18amonth
weight and score, fertility, age and progeny production. Cows were
culled at 10 years or if they failed to produce a calf for 2 consecu-
tive years. When these selection practices were applied his annual
selection differential for final test weight and 196-day gain were

14.4 and 8.5 lbs., reSpectively, which accrued to 1.46 and 1.10

standard deviations above the mean for the reSpective two traits.
Brinks selection also included a negative intensity against inbreeding
of the calf and dam which were -.20 and -.08, respectively. However,
inbreeding still increased at slightly less than 1.0% per year in the
calves and caused detrimental effects on preweaning gain and weaning
weight of bull calves. Postweaning gain was also decreased by inbreed-
ing, the reported regression coefficient was -1.6748.

Hoornbeck and Bogart (1966) reported no increases in performance
of four inbred lines of beef cattle. Because a great deal of emphasis
was placed on holding inbreeding to a minimum, selection differentials
for the performance traits were not Optimal, thus explaining in part
why no increases were monitored.

Nelms and Stratton (1967) with Wyoming Herefords were able to
obtain increases of 10.9, .05 and 3.0 kg for final weight, average
daily gain and 180-day weight, respectively, per generation. Final
weight was the trait selected for with the average selection
differential over 12 yr. being 2.41 kg per generation. Selection
differential per generation for average daily gain and 180-day weight
were .07 and 6.5 kg, reSpectively. The average generation interval
reported was 4.29 yr. Birth weight was the only trait to exceed its
selection pressure with a per generation increase of 1.2 kg compared
to its selection differential of .8 kg. One can theorize this to
happen if final weight and birth weight are genetically correlated
plus the selection differential for birth weight is secondary, thus
the expected change is not equal to heritability times secondary

selection differential.

lO

Chapman et, al. (1969) reported beneficial weight gains when
selection was practiced on weaning weight and postweaning gain in beef
cattle. Uniform selection differentials from year to year were
accomplished for performance selected herds, but when yearling type
score was the selection tool uniformity of selection differential for
performance was lost. The two herds selected for rate of gain and
weaning weight had significantly higher postweaning gain than the herd
selected on type. In a later report on the same herds, Chapman
et, 31, (1972) concluded that rate of gain herd progeny were lower
than weaning weight herd progeny in weaning weight performance, but
both were superior to progeny from the type selected herd in post-
weaning average daily gain. Correlated reSponses in birth weight,
weaning weight and yearling type score were realized when selection was
for postweaning rate of gain. When weaning weight was the selection
criterion, only birth weight reSponded in an upward manner. Chapman
and his co-workers observed that when the absolute value of the
genetic correlation was at least .4 then the direction of observed
related response and the sign of the genetic correlation were in
.agreement.

Much investigation has been done using selection to improve swine
and laboratory animals, thus only major articles will be reviewed
to bring into perSpective what selection can accomplish in other
Species. Dickerson et, 91, (1954) summarized selection differentials
and rate of performance improvement at seven state experiment stations
c00perating in the Regional Swine Breeding Laboratory. Thirty-eight

lines, which consisted of 4,521 litters, were included in the summary.

11

Inbreeding of varying degrees was practiced at most stations, thus
the average inbreeding calculated for the dams and litters was 2117
and 24:7%, reSpectively. Dickerson reported the combined overall
average l54-day weight selection differential for sires and dams from
all reporting stations, which varied in their selection procedures, to
be 14.9 lbs. The average age was 1.33 yr. for both sexes. For
weight at 56-days the combined average selection differential for both
sexes was 3.64 lbs. Although this amount of selection offered much
opportunity for improvement, none was realized when data from four
stations were analyzed. When linear regression of performance on time
was calculated the mean change for 56-day weight was -.41 lb. and
-4.0 lb. for 154-day weight. Dickerson and co-workers reported a
-3.44 lb. change in l54-day weight per ten percent increase in
inbreeding. A -2.8 lb. per year change in mean 154-day weight was
still observed when corrections for inbreeding were made, indicating
selection failed to improve genetic merit of lines being mildly inbred.
Hetzer e3, 21, (1956) reported that upward and downward selection
on backfat thickness at a live weight of 175 lbs. produced genetic
changes. In the Duroc breed after 7 generations of selection Hetzer
observed an 18% reduction in backfat for pigs selected for low backfat
and a 35% increase in backfat for those selected for increased backfat.
The control Duroc group remained at the initial backfat thickness of
1.50 in., whereas low-fat selected were 1.22 in. and high-fat selected
were 2.01 in. Hetzer (1963) also reported that the backfat selection

brought about correlated changes in feed required per pound of gain

12

and carcass cutability. The control line required 3.13 lbs. of feed
per pound of gain, whereas low-fat selected pigs consumed 3.07 lbs.
per pound of gain while high-fat selected pigs needed 3.37 lbs. Low-
fat selected pigs also showed increases in percent lean cuts while
high-fat selected pigs decreased in there carCass cutability. The
actual values for carcass cutability were 41.4, 37.0 and 39.2% for
low-fat selected, high-fat selected and control pigs, respectively,
which indeed Show that as fat thickness decreases carcass cutability
increases.

Dalton and Bywater (1963) conducted a selection experiment with
mice in which they selected the whole litter of the litters that were
high in either litter size or weight. Random mating was practiced,
except to avoid litter mate matings. From first parity data they
found no significant increase in either trait over 14 generations,
however, the litter weight selected group was consistently above the
litter size selected group and randomly selected group. In a later
experiment, Dalton (1967) found that mice on two diets reSponded
significantly to selection for growth. The mice were in a two-by-
three factorial experiment in which one-half were placed on a diluted
diet and one-half on a full-feed diet. Each diet group was Split into
thirds with the three subgroups being 1) selected for superior growth
after weaning, 2) selected for inferior growth after weaning, and
3) randomly selected for use as controls. All lines, except for the
superior growth line on dilute diet, were significantly different from

the control line within each diet. After 13 generations of selection

13

the full-fed, superior growth line was 3.0 g heavier than the full-fed
control line, 2.5 g heavier than the diluted diet, superior growth
line and averaged 6.0 g heavier than both full-fed and dilute-fed
inferior growth groups.

Rahnefeld gt, g1, (1963) in experimenting with mice found that
mass selection increased postweaning rate of gain, which was measured
from 18 to 42 days of age. After 17 generations of selection, the
estimated amount of improvement for gain was 4.9 g, which was
accomplished by total selection differentials of 36.28 g for males
and 16.14 g for females. Rahnefeld reported the 4.9 9 increase was
about six times the additive genetic standard deviation, which was
.89 g, and about 43% of the original mean growth.

Enfield gt, g1, (1966) in a selection experiment with Tribolium
Castaneum found significant differences between selected and control
groups when increase of pupa weight was the goal. Designation of
parents was on an intra-half—sib family basis with the heaviest male
and two heaviest females being selected from each half-sib family in
his S population. The control p0pu1ation, C, was composed of
individuals closest to the half-Sib family mean. The average
selection differential, after adjustment for differences in reproduc-
tive rate for the two S replicates, was 208.5 pg. ReSponse to
selection was linear over 12 generations with the regression of the
difference in mean pupa weight between S and the correSponding C on
generation time being 60.3:4.8 pg and 61.8:4.9 pg for the first and
second replicates, respectively. In 12 generations of selection no

reduction in additive genetic variance was noticed.

14

Chapman (1951) summarized the effectiveness of selection in
laboratory animals and discussed eight major points that he observed.
The laboratory experiments indicated that with or without inbreeding
selection has effectively produced changes in both the positive and
negative direction over many generations. The effectiveness of back
selection appeared to be dependent on number of generations of original
selection preceding it in some cases but not others. Chapman observed
that when crosses were made between individuals from positive and
negative selected lines intermediate offSpring were generally the
result. In general, a constant amount of variation within selected
lines has remained over the generations of selection. Correlated
reSponses were observed and in one experiment a decrease in
heritability occurred in later generations of selection. In most of
the experiments Chapman reviewed,adequate amounts of variation were
experienced,which allowed acceptable selection intensity to be
practiced. However, in one experiment all of the phenotypic variance
was due to environmental influences,which decreased the effectiveness
of selection.

In a continuous selection for egg production experiment, Dempster
gt, g1, (1952) found that with many mating systems the average flock
production appears on the surface to have improved at a decreasing rate
during the 16 years of selection. However, the second-degree regression
coefficient was not statistically significant. Furthermore, it
appeared improbable that response to selection had hit a plateau

and possible that gains in relation to selection intensity may still be

15

obtained at approximately the rate symbolic of the early years of the
improvement program.

Kyle and Chapman (1953) in looking at the effectiveness of
selection for ovarian weight in mice found that expected progress was
not obtained. The authors decided this discrepancy would disappear
if a smaller individual heritability and/or a larger litter-environment
variability were postulated. In the 14 generations of selection for
high and low ovarian response, the results were interpretted as
indicating little change in combined influence of additively genetic
and litter-environment as sources of variation. Based on unweighted
averages of all data, high groups were increased 29% and low groups
decreased 23%, reSpectively, as compared to an expected increase and
decrease of 34 and 35% for the two reSpective groups. Weighted values

were equally far apart.

 

 

Selection and Improvement jg Simulation Studies
Many studies have been done the last 17 years utilizing the Monte
- Carlo procedure developed by Fraser in the mid-1950's. In his first
and second series of papers Fraser (1957a, 1957b) discusses the logic
used by the computer to carry out the gene-by-gene simulation and how
linkage can be incorporated to cause differences. Parameters used by
Fraser were p0pu1ation Size, selection intensity and degree of linkage.
His study indicated linkage had no effect until the recombination
frequency was tighter than .025. Fraser continued to expand his Monte
Carlo approach but worked in areas that do not pertain to this

investigation.

16

In 1965, Gill reported several findings in a Monte Carlo study in
which a metric characteristic was determined by the genes at 40 loci
equally Spaced over eight chromosomes, with two alleles per locus and
equal genetic effect at all loci. Gill (1965a) reported that
restricted p0pu1ation size resulted in the accumulation of inbreeding
effects even though it was not the studies purpose. His effective
p0pu1ation sizes were 8, 12, 16 and 32 parents for which he considered
many different gene action models with three of the nine models having
conditional epistasis. Mating was done randomly.

After 20 generations the complete dominance, no epistasis model
with population of 32was significantly higher than other p0pu1ation
sizes. The p0pu1ation with eight parents after 15 generations had
regressed backwards because of inbreeding depression. With the over-
dominance model very strong inbreeding effects in the small p0pu1ations
were quick to override the positive effects of selection and also
resulted in negative regression of mean on generation number. In the
additive-by-dominance, conditional epistatic mode1,instead of inbreed-
ing causing a depression in the mean, it acted as an uplift.

Gill (1965b) in a later report pointed out the futility of predict-
ing selection reSponse based on infinite population size in graphic
p0pu1ations of realistic Size. Whether the predictions are linear or
asymptotic to the selection goal, random drift and selection have

considerable influence in changing parameter values rather quickly.

17

Gill (1965c) further reported that selection in his work was
effective in advancing the genetic mean with all models of gene
action when the optimum genotype was homozygous. However, when
the heterozygote was Optimal, selection was ineffective in small
populations under mass selection. Gill further elaborated that
major differences in genetic means should be produced with different
intensities of selection under models with a single peak of genetic
merit or one maximum fixation state. In his investigation, when
large amounts of dominance variance were present selection intensity
brought about even greater differences in the genetic means. Important
differences between actual and predicted selection differentials were
observed in populations that had no environmental variance and a gene
model involving dominance to a high degree. These two factors plus
the fixation of many loci resulted in discontinuous phenotypic
distributions after several generations of selection. Gill, in
general, detected that deviations in amount of environmental variation
between populations seem to be important in affecting the total
reSponse only when those differences existing are caused by rapid mean
change because of intense selection.

A cause of bias in estimating components of variance in small
populations was inbreeding rather than linkage disequilibrium. Gill
(1966) reported selection increases the degree of inbreeding,

especially in a population of restricted size. Linkage caused a high

18

degree of variability in degree of inbreeding, but could cause
inbreeding level to become very high.
Utilizing an additive model, Parker gt, g1, (1969) reported on

a 34

factorial simulation in which they investigated genetic correla-
tion and response to selection. The four factors which has three
levels each were genetic correlation between X and Y, intensity of
selection for X, and environmental variation of X and of Y; these
were considered in all combinations. Parker made no attempt to
control the alteration in heritability over the 30 generations,

which allowed changes in genetic parameters to occur when deviations
in environmental variation were used.

Upon analyzing the data, Parker found an interaction between
intensity of selection and heritability and a rapid decrease in genetic
correlation requiring both intense selection and high heritability.
When selection intensity became .2 genetic correlation was affected,
becoming large only when heritability of the selected trait was high.
Of course, this indicates that normal selection in our meat and milk
producing Species would cause little affect upon genetic correlations,
for most of the traits we work with are, at the most, moderately
heritable. In general, for mild selection initial genetic variance and
covariance were maintained for the entire 30 generations of selection.
Even with more intense selection genetic correlation remained near its
initial level due to the fact that both genetic covariance and variance

in the selected trait decreased together.

19

Parker reported that with truncation selection, phenotypic
correlation decreased with respect to the unselected population.
Because phenotypic correlation is a function of genetic and environ-
mental correlations between traits plus heritability, a closer analysis
had to be made to find what this reduction was due to. From graphical
displays it was learned that truncation decreased genetic correlations,
but the amount of reduction was a function of heritability rather than
degree of truncation. As heritability increased genetic correlation in
the selected offspring became less.

Using simulation methods, Ronningen (1969) experimented with
maximization of improvement with progeny testing. In order to make
progeny testing feasible as a means of genetic improvement, Ronningen
pointed out that some of the following criteria should be met: 1) low
heritability of the trait, 2) trait is sex limited, 3) trait is
measurable only after slaughter, 4) generation interval is not
increased greatly, 5) reproductive rate of the female is low and 6)
males are used intensely through artificial insemination. Also in
Ronningen's series of works he looked at progeny test stations sizes
of 500, 1,000 and 2,000, heritability levels of .1, .3 and .5, three
levels of maternal effects: 0, .l and .2, and three sizes of full-sib
groups: 1, 2 and 4. With a fixed mean, he used a random simulation
model with the following elements random: additive genetic effects,
non-additive genetic effects, permanent and temporary environmental

effects.

20

When comparing selection based on individual record only (one-
stage selection) to selection based on individual record and a progeny
test of the individual (two-stage selection),Ronningen determined a
10 to 20% advantage in two-stage selection,provided generation interval
was considered to be equal between one and two-stage methods and
number selected was kept to a moderate level. The efficacy of two-
stage selection increased as testing capacity increased; for example,
the mean genetic improvement difference of two-stage compared to one-
stage selection was 1.07, 1.10 and 1.13 for testing capacities of
500, 1,000 and 2,000, respectively. Deviations in heritability caused
little effect, eSpecially in group means. Ronningen further discussed
that genetic reSponse to selection was a function of the number
selected and that the efficacy of two-stage selection decreased as the
progeny group Size increased.

Another affect selection caused, as reported by Ronningen (1970),
was the skewing of the frequency distribution of both phenotypes and
,genotypes, with the latter being affected the greatest. Skewness
increased as both heritability and selection intensity increased. AS
heritability got larger the tail of the normal curve representing the
poorer genotypes got steeper, thus giving positive skewness. The tail
of the normal curve representing better genotypic individuals was not
affected to a significant degree by either heritability or selection
intensity.

When Ronningen (1970) compared the efficiency of combined

selection, i.e., performance testing combined with progeny testing,

21

to selection based on performance testing only the relative efficiency
of combined to performance test only decreased as heritability
increased. The average efficiencies for the heritability levels of
.1, .3, .5 and .7 were 1.40, 1.16, 1.04 and .96, reSpectively. As
expected, the greatest depression in efficiency came when heritability
changed from low to moderate status. The reasoning used is that when
heritability is low the basis for making the first selection is
inaccurate, thus the information accumulated from progeny testing is
more likely to increase the accuracy of predicting genic value.

Upon further analysis Ronningen found that the effect of common
environment iS small, but did seem to increase as heritability
decreased and when number selected got smaller. Because of this the
greatest error in ranking tested animals due to common environment
was likely to happen when trait heritability was low. Also he found
efficiency of combined selection decreased as the number of selected
individuals increased. Due to this fact, plus previous mentioned
findings, Ronningen recommended combined selection be used only when
heritability is low and intensity of selection high.

Bereskin gt, g1, (1969) conducted a simulated swine selection
experiment where two traits each were controlled by 28 pairs of genes.
TWenty of the 28 pairs controlled both traits. He utilized three
groups of which one, G-l, consisted of 16 replicates of one sire mated
to 10 dams, group two, G-2, consisted of‘8 replicates of two sires
mated to 20 dams and group three, G-3, which consisted of 4 replicates
of four sires mated to 40 dams. The heritability levels used were .12

for trait one and .34 for trait two.

22

When selection was employed on trait two, mean genotypic value
for trait two was maintained in G-l, but increased at average rates
.of .25 and .50% per generation in G-2 and G-3, reSpectively. There
was no indication of reaching a plateau by generation 10. Inbreeding,
as one would expect, increased at drastic rates for G-l, as indicated
by its coefficient of .76 at generation 11. Inbreeding for G-2 and
G-3 reached .57 and .40, respectively. Bereskin observed uneven
decreases in trait one in G-l, small decreases in G-2 for trait one
and none in G-3.

Bereskin reported that additive variance was largely maintained
because of direction of selection. Nonadditive variance was primarily
reSponsible for maintaining within variance above the expected level
in all groups for both traits. Random drift effectively increased the
amount of between variance and was apparently the major factor in
depressing and fixing the genotypes in G-l for both traits. In
general, phenotypic variation followed the same pattern of genetic
variance. In other words, the principal factor accounting for
differences between replicates within a group was random drift.

Later Bereskin (1972) reported on selection response and inbreeding
depression. Inbreeding effects were largest in replicates of G-1 and
smallest in those of G-3. For G-3, the accumulated effects from
selection over 10 generations on the growth trait was 15.49:2.33 units
while inbreeding depression was -2.04:O.10 units. The same values for
G-l were 10.68:6.4l and -9.95:1.28 units, which gave insight to what a

restricted population size can do to improvement through selection.

23

Selection effects in Bereskin's additive model were two or three times
the magnitude of any other models. Bereskin pointed out that for
traits of medium inheritance, .15 to .35, substantial long-term

genetic gains are possible even with inbreeding rates as high as in

G-2 lines,where a Six percent loss of heterozygosity occurred per
generation. His average gains were .5% or more per generation in

G-2 replicates. The reSponse was contributed to continued availability

of adequate additive genetic variance, deSpite inbreeding.

Mating Systems and Their Influence gg_PopulationS

 

 

Since the rediscovery of Mendelian genetics, there have been a
vast multitude of mating systems developed for both plants and
animals. Varied forms of inbreeding and outbreeding are the result of
mating individuals because of their consanguinity or nonconsanguinity.
P0pu1ar today is mating individuals which are either similar or
dissimilar in a phenotypic character. This system, somatic resemblance
assortative mating, can be either positive or negative, with negative
many times being called corrective mating.

Mating systems based on consanguinity and phenotypic likeness
have many differences and ramifications which will be discussed by
reviewing ground work laid by Wright and Lush. The differences
between assortative mating and consanguinity is three fold with one
being that the phenotypic likeness may not be genic. Also, except
for linkage, all gene pairs act independently of each other in
consanguine mating and the number of loci does not affect the results,

whereas in somatic assortative mating gene—by-gene Similarity iS

24

dependent on the number of loci. Lastly, somatic assortative mating
does, while consanguinity does not, produce correlations between
nonallelic genes which affect the same trait, thus affecting the
variability of the p0pu1ation. Heritability has no effect on the
genetic structure of the population in consanguine mating, but does in
assortative mating. Indeed skill of mating becomes of utmost
importance in somatic assortative mating because of results desired.

The problem with consanguine mating is the amount of inbreeding
which can accumulate and cause performance depression. Small
p0pu1ation sizes brought about consanguine mating and increases in
inbreeding coefficients in simulation work done by Bereskin (1970)
and Gill (1966). Dickerson (1949) reported close to a linear decline
in performance with increased inbreeding for different strains of
inbred hogs at experiment stations.

Nelms and Stratton (1967) found that in 302 Wyoming Hereford
calves inbreeding caused a decrease in lBO-day weight of .465 kg for
each one percent increase in inbreeding coefficient. Final weight of
the calf was not affected by inbreeding, nor did a low amount in the
dams cause a depression in any traits. Hoornbeck and Bogart (1966)
obtained results similar to Nelms and Stratton; in that, lower inbred
male and female calves had greater suckling gains. Their mildly inbred
calves had greater postweaning gain than non-inbred calves, but as
inbreeding increased there was an indication of decreased performance.
Inbreeding in the dams gave Hoornbeck and Bogart different results
than Nelms and Stratton; that is, non-inbred dams had calves with
higher suckling gains than inbred dams. Because of inbreeding

depression Hoornbeck and Bogart realized no response from selection.

25

Brinks (1965) obtained similar results from inbred lines of
Hereford cattle. The calves' and dams' average inbreeding coefficients
were 16.1 and 11.7%, respectively, with the range for each in the 26
year period being 7.2 to 26.5% for calves and 3.1 to 22.7% for the
dams. Calf and dam inbreeding had a detrimental effect on weaning
weights and inbreeding of the calf also caused a decrease in post-
weaning gain and final weight of selected bulls. For example, the
arithmetic average partial regression of weaning weight on inbreeding
of calf was -1.35 and -l.lO on inbreeding of the dam. Bull post-
weaning, 196-day gain had a -l.675 partial regression coefficient on
inbreeding of the calf and the partial regression coefficient for
final weight was ~2.296. Inbreeding of the dam was more detrimental
in its effect on weaning weight of bulls than on heifers.

Somatic assortative mating is as Falconer (1960) describes it,
when mated pairs tend to be of the same genotype more often than
would occur by chance then positive assortative mating is taking place
and if less often then the negative Option is occurring. Somatic
assortative mating can cause change in homozygosity, but because of
the large number of gene pairs controlling production traits it is
doubtful much change in homozygosity will occur. Wright (1921b)
calculated that after 15 generations of perfect assortative mating
with heritability equal to 1.0, perfect phenotypic correlation, gene
frequency of .5 at each loci and the trait controlled by 10 gene pairs,
one-third of the gene pairs would still be heterozygous as compared
to the initial frequency of one-half. Under imperfect assortative

mating (phenotypic correlation equal to .50) and other conditions not

changed from above,the value is .444 at 15 generations and infinity.

26

The effects of assortative mating to random mating on pOpulation
variance is of major importance in evaluating expected reSponse to
selection. Wright (1921b) calculated that when a trait is controlled
by 10 factors with a phenotypic correlation between assortative mates
of .75, there would be a 1.86 to 1.0 greater standard deviation than
if random mating occurred. With perfect positive correlation the
ratio is up to 4.47 to 1.0 with 10 factors. Wright further stated
that if heterozygosis has been eliminated then the genic variance
becomes equal to the number of factors controlling the trait. This
is twice the value of the genic variance in the original random-bred
pOpulation when heritability equals 1.0, regardless Of the number of
factors.

In Lush's (1948) review, the point is made that mating like to
like phenotypes tends to bring together mates whose genes have similar
effects but which need not be allelic to each other. He further
comments that similarity between phenotypes can be great while the
similarity of genotypes, gene by gene, may be drastically low. As the
number of gene pairs controlling a trait increases, less and less of
the phenotypic correlation comes from likeness in allelic genes and
more comes from the non-allelic genes.

Positive assortative mating tends to increase the frequency of
the genotypes at the two extremes and diminish the intermediate
classes. Negative assortative mating acts in the Opposite way; in that,

the intermediate classes increase in frequency and the extremes become

27

fewer. Therefore, negative tends to lower the genic variance base,
which lends to the system's other name, corrective mating.

The largest factor which determines the effectiveness of
positive assortative mating is the closeness of correlation between
the two mates genotypes. Lush (1948) explained that genetic
correlation's magnitude cannot exceed the product of heritability in
the narrow sense and phenotypic correlation between the two mates.
Furthermore, he commented that dominance, epistasis and environmental
effects tend to hold the genic correlation down because of their
depressing effect on genic effects. Even though positive assortative
mating can make a pOpulation more variable than the most inbreeding,
Lush is quick to point out the difficulty in achieving prOper
assortative mating. Moderate values of correlation have distinct
effects on the population variability. However, in order for positive
assortative mating to make a pOpulation more variable than the most
inbreeding, a heritability of .5 and a correlation larger than 2%:T’
where n is the number of gene pairs affecting a trait, must be
available.

Positive assortative mating can be practiced only on traits that
can be seen or measured. It is effective in increasing variance for
certain characteristics only if they are highly heritable and controlled
by a few genes. Li (1955) remarked in his conclusions that complete
positive assortative mating should lead to complete homozygosis of a
pOpulation but would be seldom accomplished under natural conditions

due to imperfect assorting. Li continued that somatic resemblance

28

mating may lead to population genetic compositon very different from
that reached by inbreeding. Li concluded that if breeders could
combine a system of inbreeding with positive assortative mating, the
rate of genetic fixation for a phenotypic character would be greatly
enhanced.

Experimentation with positive assortative mating in meat and
milk producing Species is not in the literature. However, a few
experiments have been done with laboratory animals, some successful
and some not. Blair gt, g1, (1961) reported highly significant
differences for each generation due to mating systems and selection in
Tribolium Castaneum. The descending rank of the mating systems for
mean larval weight was: outbreeding, phenotypic assortative mating,
random mating, phenotypic disassortative mating and inbreeding. The
only significant change in pOpulation variance was with the inbred
population. In a Tribolium study done by Wilson gt, g1, (1965), the
average reSponse of mass selected, assortatively mated lines was
slightly more than the mass selected, randomly mated lines, though this
difference was not significant. In mass selected lines, regardless
of mating system, Wilson and co-workers observed a decrease in both
phenotypic and genetic variance. Wilson gt, g1, (1966) in a later
paper reported correlations obtained with their investigation. The
phenotypic correlation between mates within line and replicate over
the generations averaged .97, -.92 and -.06 for assortative,

disassortative and random mating, reSpectively.

29

In an abdominal bristle study with DrOSOphila melanogaster,
McBride and Robertson (1963) found assortative mating to increase
the total population variance, however, not significantly. Of
the 24 lines mated in a positive assortative fashion, 18 showed
higher genetic variance. The authors creditted the increased rate
of reSponse in early generations with assortative mating to an
advantage in selection differential. McBride and Robertson further
noticed that assortative mating was best with index selected lines;
the index consisting of the animal's individual phenotype plus its
family average.

Sutherland gt, gfl, (1968) studied the effects of assortative
and disassortative mating in comparison to random mating on six week
weights in mice. Within each mating system, they practiced upward,
downward and no selection to make a 3 by 3 factorial experiment.
Their conclusions went along with previous laboratory animal work;
that is, a small and insignificant advantage in genetic improvement
was gotten for assortative over random mating. They further concluded,
as had others, that only in cases of selecting with highly heritable
characters would assortative mating be of economic advantage over
random mating. Indeed, Sutherland's experiment, plus others, point
out that the trouble of assortment may not be worthwhile, plus because
of small differences realized, no great advantages in selection
differential occurred.

Through out this section comparisons have been made between
inbreeding, somatic assortative and random mating, but no formal

definition or effects of the latter system have been given. The

3O

formal definition given both by Falconer (1960) and Li (1955) for
random mating is that, in the case of bisexual animals, any one
individual of one sex is equally like to mate with any individual
of Opposite sex. "Panmixia" is sometimes used as a synonym of
random mating.

Generally the Hardy-Weinberg Law is discussed whenever random
mating is discussed. This law points out that with random mating,
a sizable population, and no migration, mutation or selection,stability
is gained with reSpect to both gene and genotype frequencies. Also
the law points out that frequencies of genotypes in the progeny
produced by random mating among the parents are determined solely by

the gene frequencies among the parents.

Estimating Breeding Value

 

For many years breeding value estimation has been the never
ending dilemma of progressive breeders. Usually, without realizing it,
breeders tried to combine a few different types of available information
into a crude mental index and used it aS the basis of their selection.
Today, with the advent of high Speed computers, relatively complex and
correctly weighted estimates of breeding values can be made in a I
phenomenal short period of time. However, before going into past
literature and estimation methodology, definitions need to be brought
forth.

Lush (1935) described breeding value as the average effects Of
the parents' genes on the mean genotypic value of their progeny. An

individual's breeding value can be estimated from its own individual

31

record, or it can be estimated from progeny data or from a combination
of both. If the sire is mated to a random sample of the population
then his breeding value, expressed as a deviation from the mean, is
twice the mean deviation of its progeny from the population mean with
an infinite number of offSpring per sire group raised in the same
overall environment. The deviation is doubled because the individual
contributes only one-half the genes to the progeny.

Falconer (1960) further defines breeding value in terms of average
effects of genes which are the mean deviations from the population mean
of individuals which received that gene from one parent; the gene
received from the other parent having come at random from the pOpula-
tion. In these terms, Falconer (1960, p. 121) defines breeding value of
an individual as "... the sum of the average effects of the genes it
carries, the summation being made over the pair of alleles at each
locus and over all loci.“ However, such things as dominance, epistasis
and other interactions cause errors in breeding value estimation.- This
lowers the accuracy of predicting a breeding value from own phenotypic
performance or from a finite number of progeny. If the actual effects
of a gene substitution upon the phenotype are sometimes larger and
sometimes smaller than the average effect of that gene substitution in
that pOpulation, the differences between the actual and average effects
are due to epistatic or dominance deviations. If these two factors
could be deleted from effective genotypic action the job of the animal
breeder would be considerably simplified. But since dominance,

epistasis and individual environmental effects make phenotypes deviated

32

from their breeding values, procedures of estimating breeding values
from phenotypes need to be develOped so maximum accuracy of predicting
breeding values can be achieved. It is the breeding value of the
animal that influences the productivity of its offspring.

Lush, in two papers (1935 and 1947), develOped some of the early
techniques in breeding value prediction by using either individual
phenotype record, family merit or a combination of both. By use of a

path diagram, Lush (1947) found that purely family selection would

 

produce [l+(n-l)r] /*Vh[l+(n-l)t] times as rapid progress as mass
selection. The symbols used are: n for the number in the family, r
for the intraclass correlation between breeding values of members of
the same family, and t as the phenotypic counterpart of r. If t ever
becomes as large as r2 then family selection will produce less progress
than mass selection. Lush (1947) diagramatically shows the conditions
necessary for mass selection to be more effective than family
selection. Also he points out that large n improves effectiveness of
family selection, but for family selection to be more effective than
individual selection, family members must resemble each other
genetically much more than they do phenotypically.

When individual phenotype and family merit were combined into one

estimation of breeding value, Lush determined new relative effective-

 

ness 351/ “lip-3%)} film; - In analyzing this type of selection,
Lush explained that r and t must be very unequal if combination
selection is to be more effective than mass selection alone. When r
is greater than t, combining family information prOperly with

individual record will increase the rate of genetic progress.

33

Furthermore, when t is much greater than r a similar increased
effectiveness from combination selection will occur, however, in
this case family merit will be given negative attention.

In a later section, Lush (1947) develOpS equations for predicting
breeding value from individual phenotype and family average. Lush's

equation was written:

 

“ _ A l-r r-t , n A
6‘6 ' A+C+E T-t (M) + T-T THE-TIT (“38' W.

is equal to the equation's net regression on individual phenotype;
in that, A is the genic variance, C is the variance caused by
differences between overall herd effects in the whole pOpulation, and
E is variance caused by differences between individual environmental
effects within the herd. When A, C and E are totaled they are equal
to phenotypic variance. }EE, is the ratio of individual genetic
variance within families to individual phenotypic variance within
families, or the heritability of intra-family phenotypic differences.
Lush explained later that, TTTETTTE‘ is the ratio of variance among
individual genotypes to the variance among actual family averages.
The regression on family averages are automatically adjusted so that
a difference between individuals and an equal difference between
family averages will always be given importance inversely prOportional
to the variance of each. Increases of n from one to five will not
double the regression unless t is less than .38 and will not triple
it unless t is less than .17.

Lush in the same paper considered the individual as a deviation

from its family average rather than from the whole population. He did

34

this by substituting (P-Y+Y-P) for (P-P) in the previous equation and
then by separating the first two terms from the last two, expanding

and combining common terms his new prediction equation became:

9'9 = flit-E- (P'Y) “ EH (”9'

His second regression coefficient has changed to become the ratio of
additively genetic variance between family averages to phenotypic
variance between family averages or simply the heritability of
observed differences between family averages. Lush explained that in
this form family average cannot be negative, but it can be extremely
small if t exceeds r and if the genic variance is a small part of the
individual phenotypic variance. As n gets large the regression moves
toward r/t times heritability of individual differences. When the
population consists of inbred but unrelated families r/t becomes
large.

Lush (1948) summarized that conditions which make collateral
relatives most useful are: 1) there are numerous sibs to be observed,
.2) when the trait is sex limited, 3) when measurement requires
slaughtering of the individual, and 4) the trait is an all-or-none
'thing. Of course, progeny testing could do all of these things and
nu3re, but it requires an increase in generation interval.

Whether the utilization of progeny testing is feasible,
eczonomically practical and theoretically sound is of major concern.
lfle decision really to be made with progeny testing is whether to

re tain a sire or dam on the basis of their offSpring. Naturally, if

35

one analyzes in this manner, time is required and many more
individuals must have been similarly tested so selection can be
practiced. Because the individual can produce many gametes sampling
errors associated with Mendelian segregation can be reduced.

However, other factors may lower the accuracy of progeny testing.

Lush (1935) derived the correlation between the progeny average and
the parents' breeding value and later revised it to the form as it
appears in Lush (1948): er = rGPl/TTTHQTTE" Because t is composed
of both genetic correlation and other common events, one can find a
great degree of inaccuracy in progeny testing if the common event
component is large. Thus, to make progeny testing accurate, this
latter term must be kept to a minimum or be appraised and discounted
accordingly. Causes of high common event correlation,in many cases,
is the common environment the progeny are in or if progeny are by
related dams. If progeny by different sires are produced in different
years and/or in different herds, accuracy of progeny testing is
reduced. Lush (1948) concluded that if correlation from common events
rises to .25, the progeny test cannot possibly average more accurate
than mass selection. But if it is kept at zero, and heritability is
.25, data on 5 progeny are identical to mass selection. If
heritability is lower and/or number of progeny is greater selection by -

progeny testing will be more accurate than mass selection.

36

Number of progeny necessary for a highly accurate estimate of
breeding value has been a question asked many times, but one that

does not have a unique answer. In a study of Hereford progeny testing,
Stanley and McCall (1945) could not come up with the number of progeny
necessary for highly accurate estimation. However, in a similar study,
Knapp gt, g1, (1942) partially answered the question. By analysis of
variance, the pooled standard deviations between steers from the same
sire was used in Knapp's study to determine what differences in
efficiency of gain would be necessary to indicate Significant
differences in efficiency Of feed utilization. In a diagram Knapp
showed the necessary difference in efficiency of feed utilization when
number of animals in each sire group vary. The results are that the
infOrmation gained from each animal decrease drastically from 5 to 15
animals, with the sharpest decrease occurring between 5 and 10. That
is, there was a rapid increase in information gained for each animal
added, up to five, after that information gained decreased and over 15
progeny added very little to the accuracy and to what was already
known. When Knapp compared the bulls to one another by the use of
their first 8 steer progeny he found the mean difference necessary to

find significant (P$.05) differences in efficiency of feed utilization

37

was 1.52 lbs. of gain per 100 lbs. total digestible nutrients
consumed.

As mentioned previously, with the age of computers came the
age of mathematical and statistical SOphistication and animal
breeding is no exception to that rule. Many different data items
from various relatives are now commonly used to estimate breeding
values. Indeed, part of this collateral data add accuracy to the
prediction, whereas other data because of its limited quantity add
little. Willham (1973) documented the type of computerized summary
and calculations necessary to include own performance, paternal
half sibs, maternal half sibs and average progeny performance. In
Willham's example four linear equations were set up and then

converted to the matrix form shown below.

 

 

 

 

 

 

r .- .. .—
l/H 1/4 1/4 1/2 7 Bi 1
4+(N -1)H
1/4 W1]?— 0 l/8 32 1/4
4+(N -l)H
1/4 0 7N3“— (1/8 ) B3 1/4
4+ N -1 H
51/2 1/8 1/8 Walr_ £34., b1/2‘

In the left matrix the following symbolism was used: 1) N], number
of paternal half sibs; 2) N2, number of maternal half sibs; 3) N3,
number of progeny; and 4) H, heritability of the trait for which
breeding value is being estimated. The main diagonal contains the
numbers for the relatives and the off diagonals contain the relation-
ships among the relative group to the individual on the right hand
side of the equal sign. The order of the individuals on the right

hand side from tOp to bottom are the individual whose EBV is being

38

calculated, paternal half sibs, maternal half sibs, and progeny. The
values on the right hand Side are the relationships between the
individual and his relatives. If a part of the data is missing, the row
and column representing that data item are drOpped. Thus the matrix is
reduced in an upward and leftward fashion. Willham went on to explain
that the solutions to these equations are best obtained when the left
matrix is inversed and multiplied times the right hand side. Once the
values for the B's are acquired the following formula is utilized to

estimate breeding value:

EBV = B] - Individual Deviation + 82 - P.H.S. Average Deviation +

B3 - M.H.S. Average Deviation + 84 - Progeny Average Deviation .

The final value that can be calculated is accuracy, the correlation
between the animal's actual breeding values and those estimated by

the matrix process. It is calculated by taking the proper relationship
values, which are on the right hand side of the equations, times the

8 values, summing them up and taking the square root of the total, thus

 

the equation is: Accuracy=w/B] ° 1 + 82 - k + 83 - a + B4 : k . This
accuracy value simply gives a breeder an idea of how much confidence he
can place on the breeding value estimate.

Deaton and McGilliard (1965) used methods similar to ones
appearing in the last paragraph to combine and compare different
combinations of information for milk production from Holsteins. Used
in their selection index was individual record, dam's record,

daughter's record, paternal half sib records, and maternal half sib

39

records. The accuracies ranged from .50, which was for an index
utilizing only individual performance, to .68, which had individual
performance, dam's record, three daughters' records, 100 paternal
half sibs' records, and three maternal half sibs' records. The
latter index had partial regression coefficients for the records as
they were listed above of .16, .06, .19, .7 and .03, reSpectively.

After looking at many different combinations of data in
indices, Deaton and McGilliard concluded that dam's record.and
maternal half sib's records added little to accuracy of estimation.
They further concluded that paternal half sibs and daughters of a
cow could substantially increase the accuracy. In particular, the
workers noted that when number of paternal sibs was large, the
potential for increasing accuracy was greater than for any other group
of relatives.

Additional configurations were tried to evaluate the usefulness
of more distant relatives, but if the cow's individual record was
available these remote ancestors provided no information nor did they
increase the accuracy of the index. However, when no individual
record was available, the more distant relatives did have a small
value. When Deaton and McGilliard calculated the correlation of a
cow's index with an unselected daughter's record the average was .166,
as compared to .140 for the correlation of the cow's own phenotype
with her daughter's production. Thus, the index gave a 19% advantage

in genetic gain over just using phenotype alone.

4O

Nielsen (1974) investigated some features of national sire
evaluation using 100 mock sires through the use of simulation. He
looked at different Options of reference and nonreference Sires and
variable number of progeny by each sire group. Four specific data
structures were examined. Structure A had 10 reference sires and 90
nonreference sires with the latter having prOgeny in only one herd.
The number of progeny by each nonreference sire was 10 and the
number of progeny by reference sires, whose number varied from 5 to
10 in a herd, varied from 2 to 3 per sire. Structure B had 4 reference
sires and 96 nonreference sires with the latter having 10 progeny all
in one herd. Only 2 reference Sires were used per herd and the number
of progeny by each varied from 5 to 15. Structure C was the duplicate
of B except there were twice as many progeny by each Sire. Structure
0 was an undesigned structure which was used to Simulate what might
happen in field data. This data structure had more progeny per sire
than the other structures thus leading to some of the results in the
next paragraph.

The investigation also fit four data models of which one was a
simple linear model with no sire by herd interaction, two models fit
herd by sire interactions of 5 and 10% and the last model was used to
simulate data where the coefficient of variation was constant over
herds; i.e., there is more variance in a herd with a high mean than
for a herd with a low mean. Rank correlations were calculated between
the true values and the expected predicted difference values for

Sires. The approximate rank correlations between the expected

41

predicted differences and true breeding values of the sires were
.40 for A, .50 for B, .75 for C and .80 for D. In Nielsen's study
the sire evaluation is to compare herd sires against national
reference sires and against other herd sires which were compared to
the reference sires. Whereas the type of progeny test used in this

research will compare sires within a herd and not across herds.

42

Estimates gt_Genetic Parameters

Heritabilities of Traits Studied
It is in general agreement among animal breeders that the
heritability of weaning weight is low to moderate (.2 to .4). In

works done by Carter and Kincaid (1959a), Wagnon and Rollins (1959),

‘77

Shelby gt, g1, (1960), Pahnish gt, g1, (1961), Swiger gt, g1, (1962),

 

Hamann gt, g1, (1963), and Brinks gt, g1, (1964) heritability

estimates for weaning weight ranged from .43 to .69 for an average

of .51. Taylor gt, g1, (1960) and Cunningham and Henderson (1965)

'1“"‘""*‘**“

estimated the heritability of gain up to weaning time as .36 and

.50, respectively, for an average of .43. Estimates which ranged

from .30 to .36 for an average of .32 came from studies done by
Rollins and Wagnon (1956), Kincaid and Carter (1958), Blackwell

gt. g1. (1962) for heifer calves, and Minyard and Dinkel (1965).
Heritabilities which are thought to be close to the generally accepted
low level have been produced by studies done by Knapp and Woodward
(1951), Knapp and Clark (1951), Koch and Clark (1955), Brown (1958),
Swiger gt, g1, (1961), Shelby gt, g1, (1963), Brown and Gacula (1964),
and Pahnish gt, g1, (1964). Their estimates ranged from a high of
“28 to a low of .10 for a mean heritability of .21. Heritability
estimates'in a similar range for steer calves were obtained by Carter
and Kincaid (1959a), Pahnish gt. gl_. (1961) and Blackwell gt. gl_.
(1962). Kincaid and Carter (1958) found an unadjusted heritability

es timate for weaning weight to be .27. The average of all heritability

es ti mates was . 32.

43

In many studies done with laboratory animals maternal ability
has had a very Significant effect on WW and in some cases on post-
weaning gain. Two papers showing such effects in mice were Young
and Legates (1965) and Cox gt, g1, (1959). Postnatal maternal
influence was found to account for 60% of the total variance in WW
at 21-days. Hohenboken and Brinks (1971), Vesely and Robison (1971)
and Deese and Koger (1967) have shown in analyses that maternal
ability has a significantly positive effect on WW in beef calves.
Koch (1972) in his review of studies in this area points out the
correlation between milk production and gain of calves to weaning
has ranged from .3 to .8. If the median of this range is used as the
average correlation, slightly over 30% of the phenotypic variance
could be accounted for by milk production alone. Koch arrived at 29
to 33% as the amount of phenotypic variation in gain from birth
to weaning being accounted for by maternally related variation and
covariation. Koch also found in studies that genetic and permanent
environment components of maternal ability and covariance of
individual and maternal effects accounted for 35 to 45% of the
variation in gain from birth to weaning. Hohenboken and Brinks (1971)
showed in Herefords that the heritability of maternal ability was
.34 to .40.

The heritability for average daily gain after weaning is generally
accepted as being fairly high. However, Knapp and Clark (1951) when
regressing offSpring on dam found the heritability estimate to be

.18:.06. Brown and Gacula (1964), on the other hand, found that their

three different methods of estimating heritability ranged from .80

to .96. Work done in 10 other studies ranged from .39 to .76.

Carter and Kincaid (1959a) estimated heritability by three methods

of which two were consistent. When using paternal half sib
correlation estimates of .54 for females and .38 for males were
obtained and .57 for females and .40 for males were realized with
intrasire regression of offspring on dam. Regression of progeny
average on sires gave an average estimate of .21. The nine estimates
given in works done by Knapp and Clark (1951), Koch and Clark (1955),
Shelby gt, g1, (1960), Swiger gt, g1, (1961), Blackwell gt, g1, (1962),
Brinks gt, g1, (1962), Swiger gt, g1, (1962), Shelby gt, g1, (1963),
and Brinks gt, g1, (1964) averaged .51. The 13 studies reviewed had
an average heritability estimate of .51.

Not a great deal Of work has been put forth in estimation of
yearling weight heritability. Wagnon and Rollins (1959) found the
heritability of 20 month weight to be 0.44 for 305 heifers in two
herds. Blackwell gt, g1, (1962) found 18 month weight estimates of
.34 for steers and .71 for heifers. The same paper presented the
heritability of gain up to a year of age for both sexes as .32. In
work done by Shelby gt, g1. (1960), lBO-day weight and l96-day gain
were added together and then an estimate of heritability for that
figure was calculated as .55. The other studies surveyed fell between
.32 and .50. Those estimating yearling weight heritability in this
range include: Knapp and Clark (1951), Koch and Clark (1955), Swiger
gt, g1, (1961), Brinks gt, g1, (1962), Brinks gt, g1, (1964) and Brown
and Gacula (1964). The average of all yearling and long yearling weight

estimates is .46.

45

Phenotypic, Genic and Environmental Variances
of the Traits Studied

Many studies have been conducted to estimate sources of variation
in weaning weight, average daily gain and weight at approximately one
year. However, as with much research the times and length of period
for which these traits are measured are variable. Therefore, because
of this, plus wide environmental differences, much variance occurs in
the estimates occurring throughout the literature. Phenotypic
variance for weaning weight is quite variable, ranging from a low of
1271 lbs. reported by Blackwell gt, g1, (1962) to a high of 3570 lbs.
reported by Swiger and Hazel (1961). The greatest Share of works
report phenotypic variances below 2000 lbs., those included are:

Knapp and Clark (1951), Koch and Clark (1955), Pahnish gt, g1. (1961),
Blackwell gt, g1, (1962), Swiger gt, g1, (1962), Shelby gt, g1,(1963),
Brinks gt, g1, (1964) and Pahnish gt, g1, (1964) in a later report.
Brinks gt, g], (1962) and Shelby gt, g1, (1960) earlier reported
weaning variances of 2601 and 2862 lbs., reSpectively. The average
phenotypic variance in the 11 reports cited was 2070 lbs.

From the preview of heritability estimates along with phenotypic
variances, the genic variances look, as one would expect, inconsistent.
The range is from 218 to 1231 lbs., with Shelby gt, g1, (1960)
reporting the highest and Blackwell gt, g1, (1962) reporting the
lowest. The average for nine studies reviewed was 530 lbs. The non-
additive variance, which includes dominance, epistasis and environmental

variance, from the same nine studies was calculated as the difference

1’...» 9%-- - *__-—.., Mu._._.‘_-
1, A

46

between phenotypic and genic variance. When this was done the

calculated average was 1540 lbs.
The studies determining variance components for average daily

gain have not been as numerous. However, one consolation is that

most of the studies lie within a fairly narrow range when looked
at as variance in gain per day rather than variance over postweaning

feeding phase. Studies reporting phenotypic variance between .075

and .100 are: Knapp and Clark (1951), Koch and Clark (1955), Carter

and Kincaid (1959a), Shelby fl. _a_1_. (1960), Blackwell e_t_. a_1_. (1962),

Brinks gt. gl_. (1962), and Swiger and Hazel (1961). Brinks gt. g1_.

(1964) later reported a small phenotypic variance of .031 for A.D.G.

Also Shelby gt. fl. (1963) and Swiger gt. _a_1_. (1962) reported smaller

values which were .044 and .059, reSpectively. When the 10 studies

review were averaged for phenotypic variance, it equaled .073. Genic

variance reported are variable, however, most fall in the range of

.03 to .05. Investigators reporting within this range were Koch and

Clark (1955), Carter and Kincaid (1959a), Shelby gt. fl. (1960),

Brinks e_t_. _a_1_. (1962) and Swiger and Hazel (1961). The average of

these five reports was .035. Brinks gt. g1_. (1964) later reported a

value of .014 which was in agreement with a value of .021 reported

later by Shelby gt. fl. (1963). Swiger gt. g1, (1962) and Blackwell

gt. _a_1_. (1962) reported the highest genic variance values of .057

and .070, respectively. In the case of these latter two papers the

1eritabi1ity estimates were unrealistically high, thus discounting

:he validity of their genic variance estimates as being applicable to

ther beef cattle populations.

’ ! ‘au’-. m--‘-1
11

II“‘ ,

47

Scmnzinformation was found on variance estimates for yearling

mfight. Blackwell gt, g1, (1962) reported phenotypic variances of

2355 lbs. for steers and 1588 lbs. for heifers. The same author

reported genic variances of 236 and 1122 lbs. for steers and heifers,

When these two were averaged the genic variance was
Brinks gt. a_1_.

Shelby

reSpectively.
679 lbs. and the phenotypic variance was 1971 lbs.
(1964) and Koch and Clark (1955) reported similar values.

gt,qu.(l960) and Swiger and Hazel (1961) had variance estimates that

were twice as large as previous mentioned ones. Swiger and Hazel's

estimates for phenotypic variance of yearling weight were 6438 lbs.
Shelby estimated phenotypic variance at

and 3038 lbs. for genic.
In another review, Shelby

7845 lbs. and genic variance at 4315 lbs.

gt, g1, (1963) estimated the phenotypic variance for final weight at

4963 lbs. and its genic counterpart at 3172 lbs. The average of all

the studies, except for the last one mentioned, for yearling weight

was 4310 lbs. for phenotypic variance and 2082 lbs. for genic variance
Correlations Between Traits Studied

Only a few studies have made mention of the correlations which

will be used in this study. Brinks gt. gl_. (1962) and Brinks gt. g_l_.

(1964) reported two sets of correlations which are in close agreement.
In the first paper they reported correlations of .62, .67 and .59 for

phenotypic, genetic and environmental correlations, reSpectively,

between weaning weight and yearling weight. Brinks' second paper

reported values of .73, .71 and .75 for phenotypic, genetic and

48

environmental correlations, reSpectively, between weaning weight and
yearling weight. Shelby gt, g1, (1963) and Koch and Clark (1955)
reported similar values to Brinks' two sets and when averaged with
Brinks' the mean correlations for phenotype and genetic were .62 and
.71, respectively. Blackwell gt, g1, (1962) reported correlations
between weaning and 18 month-weight as .65 for phenotypic correlation,
.16 for genetic correlation and .83 for environmental correlation.

Their phenotypic, genetic and environmental correlations between

weaning weight and 18 month gain were -.10, .08 and -.16, reSpectively.

Several authors have reported correlation values between weaning
and average daily gain, however, they tend to be highly variable.

For instance, Swiger (1961) reported the high genetic correlation
between weaning weight and A.D.G. of .93 while Brinks gt, g1, (1964)
reported the low of -.20. Other authors reporting values scattered
across the whole range include: Koch and Clark (1955), Carter and
Kincaid (1959b), Blackwell gt, g1, (1962), Brinks gt, g1, (1962),
Swiger gt, g1, (1962), and Shelby gt, g1, (1963). When all eight of
the reports were average the genetic correlation averaged out to be
.33 while the phenotypic correlation average was .05.

Brinks in the two previously mentioned papers, along with Koch
and Clark (1955) reported correlations between average daily gain and
yearling weight. The average of these three papers for phenotypic,
genetic and environmental correlations were .68, .71 and .68,
respectively. Shelby gt, g1, (1963) reported higher correlations
between average daily gain and final weight off of feed test. His

phenotypic and genetic correlations were .86 and .96, reSpectively.

vfi—f "flu—L

METHODS AND PROCEDURES OF SIMULATION

Introduction

 

The actual computer process Of initiating a base parental
population, simulating their offspring and offspring from matings
in calf crops 2 through 10, storing and transferring data on local
files from one program to another, collection of population data
for analysis and soft-ware problems are indeed lengthy and complete
discussion of each phase is too detailed and lengthy for this
manuscript. Instead the programs utilized have been listed and put
in loose-leaf notebook form and stored with the author and author's
major professor. Thus, what follows in this section will be a brief
documentation of how the populations were simulated.
A total of seven FORTRAN IV programs, made Operational on the
Michigan State University CDC 6500, were utilized to accumulate
data from simulated populations to see if 1) random mating and
phenotypic positive assortative mating were significantly different
in the amount of genetic and phenotypic improvement they caused at
three heritability levels; 2) two different sire testing procedures
affect performance trait improvement; 3) different bull turnover
rates significantly affect improvement in performance traits; 4)
three contrasting heifer selection rates cause significant changes in
the amount of performance improvement; 5) three variations in selection
criteria bring about variation in trait improvement and which, if any,
excel Is in yearling weight improvement; and 6) any two-way interactions

exist between the independent variables.

49

50

BPAM and BTHS, two modified versions of the beef simulation
program written by Drs. Gordon Thomson and Richard L. Willham at

Iowa State University were used as the main programs to develop

and carry out population characteristics. BPAM and BTHS

incorporated extensive alterations of the aforementioned program

to delete unnecessary printing and supplement the output with
population parameters necessary to allow analysis of the fruitions
selection and mating programs bring about. The two versions,
BPAM and BTHS, were necessary to accomnodate both within purebred
herd testing of bulls and progeny testing of bulls within a
cormiercial test herd. This required that two selection routines
also had to be written. One selection routine was written to
accommodate within purebred herd testing of bulls and the second
acconmodated progeny testing of purebred bulls in a corrmercial test
herd while also maintaining selections within the purebred herd
itself. A more detailed description of these two routines will
follow in a later section.

The investigation was setup to use random mating (RM) and
positive assortative mating (PAM) in the purebred herd and RM in the
progeny testing of bulls within a conmercial test herd. In order to
accomplish this two RM and one PAM programs had to be made operational.

A brief description of each will follow in later sections.
All seven programs were compiled and stored on disk permanent

file and on magnetic tape to accomplish easy access by control cards.

 

 

and. f. H... h31.. 3;}..11‘. ‘ .fi.

21!

. “I, was
. ..

 

51

The flow chart in Figure 1 shows how the different programs were
used when alternate Options were chosen in Simulating herds. Only
one herd, a purebred herd, was initiated if the option of not
progeny testing within a commercial test herd was chosen, whereas
two herds, a purebred and a commercial test herd were initiated for
the other case. After initiation of the herd was completed,
selections and matings within the herd's population took place and
were passed on to either BPAM or BTHS. The matings were carried
out, offspring developed and information for selection supplied to
the selection routines. The cycle of the three programs took place

until nine selections and matings occurred.

52

FIGURE 1. FLOW CHART OF PROGRAMS UTILIZED TO SIMULATE BEEF HERD
POPULATIONS.

 
     
 

IRE
ESTING
WITHIN
A COMMERCIAL
HERD

 
 
 

 

  

YES
STORED
DATA

 
   

 

 

 

 

 

 

  

BASE PARENT

INITIATION STORED P.8. & COMM.

DATA HERD PARENT
INITIATION

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

c(/’R R
I ‘f :EEEI}

 

 

 

 

 

 

 

 

 

 

PAS I THS
R

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

R
R BTHS
DATA
BPAM
DATA
"° “5 R Q
4 y ' R
I THMTG
‘7 A R " ‘7
RNMTG PAMTG
1R IR

 

 

53

With the use of control cards, the CDC 6500 created local files,
P-2 through P-4, which contained parameter data that allowed the prOgrams
to function prOperly for specific types of data collection. If
further information on parameter files or other description is needed,
the program notebook should be consulted. The stored data file was
made into a permanent file on disk and added to each time the
simulation run was made. The other data files were local files which
acted as go-betweens for the programs as they functioned in sequence.
An R by the arrows going to and coming from the local files indicate
they were rewound before being either written on or read from.

In the next section, Simulation of Herd Population, values which
give a heritability of 40% for yearling weight (YW) will be used to
show how base parents are initiated, their phenotypes formed and
resulting offSpring of a mating generated. These values plus the ones
for the other heritability levels and how they are obtained will be
discussed in the materials section. To present a numerical example
in the next section, heritabilities of 19.7 and 59.1% for weaning
weight (WW) and average daily gain (ADG), respectively, were used to
make YW heritability 40%. The definitions and values, if they have
any, for the abbreviations used in the next section are given in

Table 1.

Simulation gt_Herd P0pulations

 

BPAM and BTHS had limitations which fixed maximum herd size'

at 90 cows and 9 bulls. In the initial calf crop 90 base

 

54

cows and 9 base bulls were created for each herd. Then 10 cows were

Inated to each bull to produce 90 Offspring. The biological model

followed for the production of each offspring is found in Figure 2.
The first calf crOp production is the only time when parents are
created, after that the individuals are created by simulating the
mating of two individuals and the biological happenings diagrammed‘
in Figure 2.

Throughout the 10 calf crops the cow herd size in both the
purebred and commercial test herds was kept at a constant size of
90. The average death rate for cows and bulls was set at 5%.
Determination of death in the cows, bulls and their offSpring was by
random number drawing. On the average 10% of the offSpring died.

Sex of the calves was also determined by random number drawing with

50% being each sex on the average.

As shown in Figure 2, base parents of the purebred herd and, when

in use, commercial test herd had three breeding values created and

stored with their herd identification. The three breeding values

created by use of multiplication factors and standardized random

normal deviates were genetic abilities for WW, ADG and maternal ability

(MA). Also a permanent environmental effect for MA was created and

stored for each female. The program was setup with no genetic

correlation between WW and MA.

Formation of genotypes and phenotypes for base parents in the

BPAM and BTHS was accomplished in the same manner for both

sexes. Equations 1-3 show the creation of genotypes for WW, MA and

55

TABLE 1. ABBREVIATIONS, VALUES AND DEFINITIONS OF FACTORS USED IN
FORMING GENOTYPES AND PHENOTYPES FOR SIMULATION OF HERD
POPULATIONS SECTION.

 

 

Abbreviation Valuea Definition
G . , G , , G , Genic value for WW in dam, sire and
D D D S D 0 offSpring, reSpectively
GM.D’ GM.S’ GM-O Genic value for maternal ability in rd
dam, sire and offSpring, respectively 1
GF-D’ GF-S’ GF-O Genic value for ADG in dam, sire and i
offspring, reSpectively g
P . , P . , P , Phenotypic value for WW for dam, sire i
N D N S W 0 and offspring, reSpectively !
PF-D’ PF-S’ PF-O Phenotypic value for ADG for dam ,
sire and offSpring, respectively
PY-D’ PY-S’ PY-O Phenotypic value for YW for dam,
Sire and offSpring, respectively
PM-D Phenotypic value for dams' maternal
ability
SD, SF, SM Genic Mendelian segregation for WW,
ADG and maternal ability, reSpect-
ively
E(W), E(F), E(M) ' Random environmental effect on WW,
A06 and maternal ability
E(P) PESDW'RN(i) Permanent environment on dams'
maternal ability
CWGlb .04802 Factor required to produce genetic
correlation between WW and ADG
CWGZb .18600 Factor required for total genic
variance in ADG
CWW 21.21320 Factor required for genic variance
in maternal ability
DGSDW 22.19234 Factor required for direct genic

variance in WW

56

TABLE 1. -- Continued

 

 

Abbreviation Val uea Definition
DESDW 26.11116 Factor required for direct envir-
onmental variance in WW
,2
DESDG .16000 Factor required for direct envir- '
omental variance in ADG 1
i
FESDW 27.52358 Factor required for direct envir- i
onmental variance in maternal E
effect on WW 1
PESDW 15.07528 Factor required for permanent l
environmental variance in if
maternal effect on WW
HEW Common herd environment for WW
for one calf crOp, equal to 10%
of standardized random normal
deviate times DESDW
HEF Common herd environment for ADG
for one calf crop, equal to 10%
of standardized random normal
deviate times DESDG
MWW 450 Mean value used for WW
MADG 2.50 Mean value used for ADG
NDGT 160 Number of days males and females
are gain tested
RN(i) Standardized random normal deviate

(SRND) used to develop equations
on pages

 

aIf a value appears it iS one necessary for a YW heritability of

40%.
b

See Appendix A for calculation and explanation; both CWGl and

CWG2 are used both base parent initiation and calf crops 1 through 10.

57

a ”WNW
OoDw
:—

.F open» cw Leeann mcouumw so meowuwcwemu use meowumm>mgnn<m

a:

9.2

a/ new
2: ._
new
93,. 3:

 

AI11.om
on: All oakalAmm
m:
mew
:3
mac ._

m.m4<sz< ouh<4=2~m 2H mHH<mH
moz<2mou¢ma azHozmsgqu machu<u mzh no z<xw<~a .N mmszu

58

ADG, respectively, for dams. Equation 4 shows the creation of the

(1) 30.0 = 22.19234 - RN(l)
(2) GM,D = 21.21320 - RN(2)
(3) GF.D = .04802 - RN(l) + (.186000 - RN(3))

(4) E(P) = 15.07528 - RN(4)

(5) Pwoo = 450 + GD,D + (26.11116 - RN(5)) + (27.52358 - RN(6)) f
(6) PF-D = 2.50 + GF,D + (.16000 . RN (7)) 1
(7) PY,D = Pw,D + (160 - PF D) i

permanent environmental effect on MA. Equations 5-7 Show the calcula-

'K'Lnlj

tions necessary to arrive at phenotypic expression for WW, ADG and YW,
reSpectively, for dams. For all cases, the abbreviations and values
are used as listed in Table l. The correlation between GD-D and GF-D
of .25 was produced through the use of the same standardized random
normal deviate in equations 1 and 3. Along with the genotypes and
phenotypes being stored, how each parent deviates from the herd mean
for YW is stored. This individual deviation (IR) is used in a later
program segment to calculate the estimated breeding value for YW
(YW-EBV).

The procedure used to create animals after the base generation
is shown in Figure 2. Half of the offSpringS' genic value comes from
each parent with apprOpriate Mendelian segregation being added.
Because this path diagram develOpS the phenotype for only one
individual, overall herd environmental effects on individual WW and

ADG were omitted. As shown in Table 1, the common environment for

59

each individual in a herd-year group is 10% of the magnitude of direct

environmental standard deviation for the trait times a standardized

random normal deviate. There is one random draw for each calf crOp.
The actual formulas used to calculate offsprings' genotypes and

phenotypes are represented by equations 8-13.

(8) e (60., +

0-0 GD.D)/2 + (/‘.‘5- 22.19234 - 1211(1)) ..

(9) GM.0 = (GM.S + GM.D)/2 + (/‘i§'- 21.21320 - RN(2))

 

(10) GF.0 = (6F,S + 6F,D)/2 + [v575'(.04802 - RN(l) +
.186000 - RN(3))] i
0 + (26.11116 - RN(5)) + GM.D + E(P) +
(27.52358 - RN(6)) + HEW
+ (.16000 - RN(7)) + HEF

 

(11) P 450 + G

W‘O D:

(12) PF,0 = 2.50 + 6F,0

(I3) PY'O = PN‘O + (160 ' PF’O)

These formulas were utilized in forming progeny in all calf crOps.
BPAM and BTHS were finitely developed to summarize and

store data needed for YW-EBV calculations as described

by Willham (1973). In particular, for the YW-EBV calculation the

following four data items are used: (1) own performance expressed as

a deviation from contemporary group (IR), (2) average performance

of PHS as the average of the individual deviations and number of PHS,

excluding the individual under consideration, (3) average performance

of MHS as the average of the individual deviations and number of MHS,

excluding the individual under consideration, and (4) average

60

performance of the individual's progeny expressed as the average of
the individual deviations and the number of them. Due to programing

difficulties PHS and MHS records are never updated once they are

established. Indeed, added accuracy could be gained from more PHS

data, but one progeny record adds to accuracy at twice the rate a

PHS record does thus making half sib data update less essential.

Progeny average was re-evaluated every time the individual produced

b ‘1 Y‘fi.."r’iI
.

another offspring. All items used in YW-EBV were expressed as

deviations from the mean of the group in which they were produced. If

War a

a data item was missing then the row and column correSponding to that

e n...

data item were drOpped out of the simultaneous equations for YW-EBV.
For instance, if only IR and PHS records were available for an
individual the rows and columns corresponding to MHS and progeny

average were drOpped and only two 8 values, as described by Willham

(1973), were estimated. The resulting YW-EBV would then be the two

8 values multiplied by their respective independent variables, IR and

PHS.
BPAM and BTHS incorporated numerous modifications which allowed them

to read from and write on local files in the computer were numerous.

As diagramned in Figure 1, information had to be passed onto the

selection programs so correct selections could be made and the selected

animals passed onto the mating programs. BPAM and BTHS wrote onto a local

file the identification number, ADG and YW phenotypes ”and YW-EBV'for the

calves. In the case of individuals which are in production, progeny

61

averages for ADG and YW plus YW-EBV were also written onto the local

file. Rewinding (R in Figure 1) of the local file was necessary before

utilization by the selection programs could occur. In the case of

reading the local file formed by mating programs, the BPAM. and BTHS

were altered to read the local file as card form images which the Thomson-

Willham program currently utilizes for most classroom work.

Mechanics g_f_ Selection Prgcyams

Two selection programs were developed to work in coordination

’h-Q ' -ufi 'v
I

with BPAM and BTHS. One selection program carried out the correct

selections when only the purebred herd was used, while the other

program carried out all the selections for the purebred and the

conmercial test herd when it was in use. Both selection programs used

individual animal information read Off the local file created by

either BPAM or BTHS as shown in Figure 1, plus information from the

parameter local file, P-3. P-3 specified which trait to select on and

how many of each sex and age group were to be selected.
Selection of bull and heifer calves in the purebred herd when the

comnerci a1 test herd was not utilized was based on either phenotypic

expression for ADG or YW or the YW-EBV. Mature bull and cow selection

in this instance was based on progeny records for ADG or YW or on
YW-EBV which was revised each time individuals produced more progeny.

The numbers selected will be discussed in the Parameter Values Used

section.
When the commercial test herd was used, the selection program, test

herd selection (THS), carried out more functions than the other selection

62

program, PAS. As with PAS, THS read in all sex and age classifica-

tions and ranked them via FORTRAN IV double do-loops based on? the .

selection criteria read from P-3. Both THS and PAS selected heifer

calves for the purebred herd based on their own performance records.
These heifers were complemented with the highest ranking mature cows
based on their progeny records or YW-EBV to make the breeding herd if -1
size 90 head. THS also kept the commercial test herd at 90 cows '

by replacing the dead mature cows with the highest yearling weight

heifers. .
Bull selections made by THS were more numerous and complicated E

 

in comparison to those described previously. Two stage selection

took place before a bull was used in the purebred herd. The first of
the two stages of selection came when the top 9 bull calves of the
purebred calf crOp were selected based on their own performance record

or YW-EBV. These 9 performance tested bull calves were then randomly

mated to conmerci a1 cows and progeny evaluated. After progeny

evaluation in the Commercial test herd THS selected either 3, 4 or 6

of the 9 performance tested, progeny evaluated bulls to be used in the

purebred herd. Selection of the performance tested, progeny evaluated

bulls for the purebred herd was based solely on progeny average, except

when YW-EBV was the tool used for selection. In this latter case,

individual deviation from the herd mean for yearling weight, plus
paternal and maternal half-sib performance and commercial progeny

average yearling weights were incorporated into the breeding value

estimate.

63

For two of the three bull selection options, selection of a
portion of the bulls for the purebred herd was from bulls that had
previously been used in the purebred herd. Therefore, the selection
of these bulls was three stage. Firstly, they were originally
selected on their own performance record or an estimate of their

yearling weight breeding value before progeny evaluation in the

«a!!!

commercial test herd took place. Secondly, the bulls were selected

.gm....

for use in the purebred herd based on progeny evaluation as
previously described and thirdly, the bulls were selected for reuse
in the purebred herd based on further progeny evaluation which had

taken place in the purebred herd.

Mechanics gt_Mating Programs

 

Three mating programs were developed to accomplish the deeds of
this investigation. RNMTG was a RM program utilized in the purebred
herd only portion of the investigation. The program used in all cases
where the design required PAM was PAMTG. THMTG was another RM program
which besides RM the purebred herd also RM the 9 bull calves from the
purebred herd with the 90 commercial test cows.

RNMTG and THMTG accomplished RM in the same manner, except that
THMTG had to also RM the 9 purebred bull calves to 10 commercial test
cows each. When the males and females were read in from the rewound
local file created by the selection programs they were placed in two
arrays, one for selected males and one for selected females. Each
male and female was then assigned a value from the random number
generator function within the computer. A FORTRAN IV double do-loop

was then employed to rank the arrays in descending order according

64

to the assigned random numbers. In the purebred herd, the bull with
the highest random number was assigned the 15 cows with the 15 highest
random numbers, the bull with the second highest random number was
assigned cows which ranked 16th through 30th by assigned randon numbers
and so on until cows ranking 76th through 90th were assigned to the
bull with the smallest random number. The same tactics were used to f"‘
accomplish RM of the 9 purebred bull calves to 10 cows each in the 1
commercial test herd.

Accomplishment of PAM in the purebred herd was not as easy. F
Assorting of the animals was based on their own phenotypic expression. (
Even though the mature cows and bulls were selected on their progeny
averages they were PAM on their own phenotypes as were bull and heifer
calves, thus phenotypes had to be transferred with each individuals'
identification number from the selction program to PAMTG. When PAMTG
read in data from the local file created by PAS it created two arrays,
one for males and one fOr females. Then based on mating criteria
(i.e., whether to PAM on ADG, YW or YW-EBV as read from the local
file P-4) PAMTG ranked, via double do-loops, the male and female
arrays in descending order. PAMTG then assigned the first male
with the first 15 females, the second ranking male with the second
15 females and so on until all 90 females had been assigned with the

6 selected males.

PARAMETER VALUES USED

Many objectives are meant to be evaluated in this investigation,
however, the primary objective of this study is to evaluate the
reSponse of simulated pOpulations to two sire testing procedures and
to two mating systems. Two sire evaluation systems were chosen which

would closely resemble what a moderate sized purebred beef herd

..N~.-i)
1

might be able to accomplish. The first was an evaluation of sires
within the purebred herd, while the second system utilized a

commercial test herd to progeny test 9 bull calves before their use

{thwart-_T '7’" 4

in the purebred herd. Naturally, the latter system requires a longer
generation interval plus more capital investment in the way of 90
commercial cows.

The two mating systems under investigation are random and positive
assortative mating. These were used as explained in the Method of
Simulation section. It should be mentioned again, however, that only
random mating was used when progeny testing of bulls was done in the
commercial test herd.

Four other items were factorialized within the two sire evaluations
and two mating systems. Three levels each of YW heritability and
percent heifer calves saved were looked at,along with three combinations
of young and old bulls saved plus selecting on either ADG, YW or YW-EBV.
A discussion of the YW heritabilities will follow in a later section.

The three levels of heifer calves saved were 20, 50 and 80%.

These three levels were chosen because they approximate the full range

of selection intensities being used today plus these levels will

65

66

hOpefully show advantages and disadvantages of generation turnover
and selection intensity.

A total of 6 bulls was always used in the purebred herd, but
three different combinations of young and old bulls were used to
acquire this total. The three combinations were 0 old and 6 young,
2 old and 4 young, and 3 old and 3 young. Of course, the first of
the three was used to find the effect maximum sire turnover rate
might have on rate of improvement. The other two combinations were
chosen to find what combination of young and old sire numbers would
Optimize selection intensity and accuracy, thus lending to maximum
rate of improvement.

The selection traits in this project were chosen because of
their interest to the author and his committee, plus large emphasis
is currently being given to them by the purebred beef industry. In
addition, it was of interest to investigate the impact YW-EBV would
have on genetic improvement and how it would compare to other
selection programs. As stated before the three data items used as
selection criteria were ADG, YW and YW-EBV. ‘Selection in the old
bulls and mature cows for ADG and YW is based on progeny averages for
the two traits. Whereas, selection of the bull and heifer calves is
based on their own phenotype for the two traits. The only exception
to this is when the bull calves are progeny tested in the commercial
herd at which time they are selected on their progeny averages.
Selection by YW-EBV calculation is accomplished in the same manner
regardless of sex or age category,except for bulls progeny tested in

the commercial herd. Because of great difficulty encountered in

67

transfer of data from program to program only the tested bull
calves' individual deviation from herd contemporaries and average
deviations of progeny was used in the YW-EBV calculation. The
YW-EBV calculation for all other individuals utilized half sib data
in addition to IR and average progeny deviation when it was available.

Three levels of YW heritability were chosen to cover the range
of heritabilities cited in the literature. The three levels used in
the simulation were 20, 40 and 60%. In developing the variance
components of WW and ADG which would give the three levels of YW
heritability, certain maximum values and ratios between various
components were standardized. The total phenotypic variances for WW
and gain from weaning to yearling time was set at 2500 and 1600,
reSpectively. The latter, when put on a gain per day basis, had a
phenotypic variance of .0625. Also set out was that heritability of
ADG would not exceed 80%. The ratio between WW and ADG heritability
would be a constant 1:3 ratio as long as heritability of ADG remained
equal to or under 80%. But when YW heritability was set at 60%, the
1:3 ratio caused the 80% limit for ADG heritability to be exceeded,
thus a 1:2 ratio was used.

As stated in a previous section, the genetic correlation between
WW and ADG was set at .25. This was kept constant across the three
YW heritability levels. Of course, the genic and phenotypic
variances for YW were the summation of genic and phenotypic variances
for WW and ADG plus two times the genic covariance between WW and

ADG. The following two equations depict genic (l) and phenotypic

68

(2) variance, respectively, for YW, where

2 - 2 2 . 2 . O
2 _ 2 2 , 2 . .

2 2 . . . 2 2
CG” and 6Pw are gen1c and phenotyp1c var1ances for WW, OGF and °PF are

genic and phenotypic variances for ADG and GGWGF is the genic covariance
between WW and ADG. Naturally, as heritability of the traits increase
so does the genic and phenotypic variance due to increased covariance.

PrOper heritability levels for WW and ADG, which give the
prescribed YW heritabilities, were gotten by trial and error method.
The calculated heritability levels of WW and ADG, reSpectively, for YW
heritabilities of 20, 40 and 60% were, 9.45 and 28.35%, 19.70 and
59.10%, and 39.25 and 78.50%, respectively. The expected genic and
phenotypic variances, reSpectively, for the three heritability levels
of 20, 40 and 60% were 853 and 4264,_l779 and 4441, and 2792 and'
4665, respectively.

Once the prOper WW and ADG heritabilities were determined it was
decided to maintain constant ratios between the non-additive variance
components in WW, for the direct genic variance for WW would vary a
great deal over the different WW heritabilities. The six ratios
between the other four variance components were as follows: (1)
individual environmental to maternal genic, 2:1, (2) individual
environmental to maternal permanent environmental, 3:1, (3) individual
environmental to maternal temporary environmental, .9:1, (4) maternal

genic to maternal permanent environmental, 1.4:1, (5) maternal genic

69

 

 

 

 

TABLE 2. COMPONENTS REQUIRED TO GIVE PRESCRIBED LEVELS OF YEARLING
WEIGHT HERITABILITY.
YW Heritability
20% 40% 60%
Componenta Value Percentb Value Percentb _ Value Percentb
DGSDW 15.37043 9.45C 22.19234 19.70C 31.32491 39.25C
DESDW 27.72762 30.75 26.11116 27.27 22.71127 20.63
FGSDWd 19.60639 15.38 18.46337 13.64 16.05930 10.32
FESDW 29.22748 34.17 27.52358 30.30 23.93978 22.92
PESDW 16.00855 10.25 15.07528 9.09 13.11236_ 6.88
DGSDGe .13304 28.35f .19209 59.10f .22159 78.50f
DESDG .21166 71.65 .16000 40.90 .11576 21.50
CWW 21.21320 21.21320 21.21320
CWG19 .03326 .04802 .05540
cwczh .12880 .18600 .21455
aAbbreviations and definitions of compOnents are given in Table l.

bPercent of total phenotypic variance.
cHeritability in percent for WW.

dAmount of WW variance due to genic variance in MA, value not
used in BPAM or BTHS. '

eAmount of ADG variance due to genic variance, value not used in
BPAM or BTHS.

fHeritability of ADG.

gFactor required to give genetic correlation between WW and
ADG (See Appendix A).

hFactor required to give prescribed genic variance in ADG
(See Appendix A).

70

to maternal temporary environmental, .4:1, and (6) maternal permanent
environmental to maternal temporary environmental, .3:1. The
components with ratios to individual environmental variance (IE) were
setup in the following equation to equal non-additive variance:

IE + IE/2 + IE/3 + lO-IE/9 = Non-Additive Variance.
The other components were found by using the ratios setup between
them and IE. With these ratios the prOper weightings for each variance
component of WW was established and values calculated to obtain each
level Of YW heritability. The figures listed in Table 2 are multipli-
cative factors which were used in BPAM and BTHS programs
to give the prescribed variances and heritabilities. The procedure

for calculating CWGl and CWGZ is given in Appendix A.

RESULTS AND DISCUSSION

The data presented in this section are from several analyses done
on the calf crOp information stored by the two modified versions of
the Thomson—Willham program. The calf crop information, because of its
volume, was stored on disk permanent file (and backed up with a data
file on magnetic tape) for ease of handling, manipulation and
incorporation into the Michigan State University computerized
statistical software package. Through the use of short FORTRAN IV
programs the large arrays of calf crOp information were broken down
to smaller data arrays, which were made into permanent files, and
statictically analyzed. Analysis of variance procedure was done in
two parts. One analysis procedure analyzed the data comparing the two
mating systems, RM and PAM, and their four factorialized sources of
variance, while the second analysis procedure analyzed the two sire
testing procedures and their four factorialized sources of variance.
The four factorialized sources of variance in both analyses were three
levels of YW heritability, three bull combinations, three levels of
percent heifer calves saved (PHCS) and three different selection
methods. The magnitude of the two-way interactions was also evaluated

in the analyses procedures.

The Effect gt_Main Sources gt_Variance

 

 

Mating Systems

Effect 9g Yearling Weight

 

Random mating (RM) and positive assortative mating (PAM), as used

in this investigation, were not significantly different in the amount

71

72

of improvement they caused in YW over ten calf crOps. Table 3 shows
that RM populations were better than PAM pOpulations for average calf
YW in calf crop 4 (CC4). However, at calf crop 7 (CC7) and calf crop
10 (CClO) PAM populations averaged higher than RM populations. In the
latter two calf crops, average progeny yearling weight for sires and
dams was higher for PAM than RM. At CClO the sires averaged 4.7 lbs.
heavier in their progeny average for YW while dams averaged 4.1 lbs.
heavier in their progeny average for YW with PAM, however, neither
mating system was significantly different from the other for these

two calculations.

PAM brought about an insignificant increase in phenotypic variance
for calf YW. The probable reason why PAM does not increase the amount of
genic or phenotypic variance is that the lower portion of the phenotypic
distribution is continuously culled off. In CClO phenotypic YW variance
for PAM populations averaged 4528, while RM populations averaged 4336.
The average expected phenotypic and genic variances for YW were 4457 and
1808, respectively. The difference in genic variance for calf YW was
smaller, with PAM and RM variances equal to 2253 and 2132, respectively,
at CClO. ‘Because genic values and environment were not correlated one
would not expect the difference between genic and phenotypic variances for
the two mating systems to differ. However, chance alone could cause this
much variation to occur between differences in the two mating systems
fOr genic and phenotypic variance. Even though a highly variable
population is not completely desirable, if the mating system creates
greater variance in the breeding population more individuals in the
IJpper extreme may be produced, thus enabling incorporation of animals
into the breeding population which average higher in genic value. Also

'the breeder may be able to propagate a product of greater breeding

73

 

 

N.ooo oo_.o o._No o ooo oNo.N o.Noo o.Poo oNN.N o.Noo oo<
N.ooo ooo.N o. _No N.omo o_o.N P.ooo N.Noo oNo.N m.oNo 3» a_eaoozo
~.ooo NNo.N N. ooo o.ooo ooo.N o.ooo o.__o oFN.N N.oNo ooo
.oo. mooo.v mooo. v mooo.v oooo.v mooo.v ooo. mooo.v mooo.v .o_m ootouopao
N.o_o_ oo,.o o. o_o N. ooo opo. N o.moo o. N_o ooN.N N.oNo Noo
m.ooo_ oNo.o o. Noo o. Noo N_o. N o.ooo o. o_o ooN.N o.NNo Noo oo>om
o.omo ooo.N o. ooo N. NNo opo. N o.oNo o.ooo ooo.N o.ooo NoN mtooooo
mooo.v oooo.v mooo.v mooo.v oooo.v oooo.v mooo.v mooo.v oooo.v .oom ooooamo
o.Noo _oo.o o.ooo o. Noo ooo.N o.ooo N.ooo o_N.N N.oNo o\o
o.ooo ooo.o o.ooo o. ooo moo.N o.ooo o.ooo NoN.N o._No o\N oo_oooooeoo
o.ooo oNo.o o._oo _. ooo oNo.N o.ooo o.Noo oNN.N o._No o\o
ooo. ooo. ooo. oNo. Soon oNN. oNo. Foo. ooo. .opm e_oo
N.ooo_ ooN.o o. oNo _.ooo oNo.o o.ooo o. oNo ooo.N o.oNo Noo
o._oo _No.o o. ooo N.Noo ooo.N N.Noo o. ooo ooN.N P.oNo Noo sooo_ooooao=
o.oNo _oN.N N. Noo o.ooo o_N.N N.¢No o. ooo oNo.N P.ooo NoN
mooo.v oooo.v mooo.v oooo.v mooo.v oooo.v oooo.v oooo.v oooo.v .oom 3>
N._oo moo.o N._oo o.ooo Noo.N o.ooo o.moo ,mN.N N.ooo zoo
o.ooo NNo.o o.ooo o.Noo ooo.N o.ooo N.Noo N.N.N P.oNo so zoomso
ope. moo. ooo. ooN. o_o. ooo. ooo. _o_. moo. .o_m oooooz
3> oo< 33 3» oo< 33 z> oo< 33 sgoooooo ooaoom
o_ ooco a_oo N ooao e_oo o oozo a_oo

 

.3> oz< oo< .33

mom <H<o 2mhm>m wth<z 2H moz<~m<> no mmumaom.

zH<z mom mz<mz mo hmH4 oz< moz<~m<> no mHm>4<z<

.m m4m<h

74

the breeder may be able to propagate a product of greater breeding
value and possibly greater monetary value. However, these data do not
support the theoretical ground work laid out in the review of the
literature. BecauSe PAM did not significantly stretch out the
variance, increased selection intensity for YW was not possible. Due
to this fact, significant increases in the improvement of YW were not
made with PAM. Thus, as an adviser to breeders, one would not be
warranted in suggesting the use of PAM, as used in the investigation,
as a tool to accomplish significantly greater genic improvement in YW.
One would have to state that the additional time and expense required
to assortatively mate would not be rewarded economically with extra
pounds Of calf at a year.

When the correlations between mated individuals were analyzed
and two-way interaction tables compiled, the phenotypic correlation
between mates for YW selection and PAM were highly positive (+.88) and
significantly (P<.Ol) different from the same correlation for RM,
which was -.06. Indeed these two correlations would make one believe
greater gains in YW improvement might be accomplished. However, when
the genic correlations between mates for YW were analyzed the
differences, although significant (P<.Ol), were small, .16 for PAM vs.
-.06 for RM, thus again explaining why PAM did not give additional
improvement in YW.

In retrOSpect, one would say PAM looks exciting in theory, but
\vhen it was actually practiced in this investigation additional rewards

I \Nere not harvested, thus supressing ones enthusiasm towards it.

75

Effect gg_Weaning Weight

 

The component parts of YW, WW and 160-day gain, reacted in
Opposite ways to the two mating systems. As presented in Table 3,

a phenotypic significant difference in WW existed in CC4 and CC7
between the two mating systems. There was, however, no genic
difference between RM and PAM in the three calf crOps analyzed. Sire
and dam direct genic values for WW were in total agreement with what
was found for direct genic values in calves; in that, there were no
significant differences found between RM and PAM. When the phenotypic
and genic variances for WW were analyzed no significant differences
between the two mating systems were found. In fact, RM caused more
phenotypic variance in WW at CClO (2506 vs. 2475), however, PAM caused
slightly more genic variance (534 vs. 530) in WW at CClO.

When genic maternal ability for calves and parents were analyzed
no significant differences were found between mating systems. How-
ever, the analysis did reveal that RM pOpulations were continuously
higher in direct genic value on WW and genic maternal ability in both
the calves and parents, but that these differences narrowed as time
elapsed.

Indeed to recommend PAM for greater improvement in WW would be one
Of poor judgment. Even though no direct selection was made on WW in
this investigation, it is relatively safe to assume that PAM would not
give additional improvement in WW. Indeed phenotypic correlation
between mates will be high (r>.85), but the lower heritability of WW

will cause lower genetic correlations to exist between mates, thus

76

lowering even further the possibility of more improvement through the
use of PAM. Due to this fact, an even lower frequency of best-to-best

matings would take place.

Effects gg_Average Daily Gain

 

Average daily gain for l60-days reacted more favorably to PAM than
RM. AS Table 3 Shows PAM was significantly higher in genic ability and
phenotype for ADG in CC7 and CClO. Sire and dam genic abilities for
ADG followed the same significant pattern of those for calves. In CClO
the sires averaged .041 lbs. per day higher for PAM than RM and dams
likewise, only the difference was .038 lbs. per day.

Because ADG was more highly heritable, the theoretical concepts
of PAM took place. In CC4 and CClO PAM caused significantly greater
variance in both genic value and phenotype. AT CClO PAM had a 10.7%
greater phenotypic variance and a 18.6% greater genic variance.
Indeed these increases in variance gave PAM an advantage in selection
differential. PAM was also substantially and significantly higher in
both genetic and phenotypic correlation between mates in the three
calf crops analyzed. When ADG selection in the two mating systems was
compared very high phenotypic correlations existed between mates for
ADG. The average phenotypic and genetic correlation between mates for
the three calf crOps for ADG selection in PAM were .90 and .30,
reSpectively. The same two correlations for RM were .02 and .01,
reSpectively. Because ADG was consistently higher in its heritability,
greater genetic correlations occurred, thus allowing a higher frequency

of best-to-best matings which acted to further stretch out the

population variance.

77

Effect gg_Estimated Breeding Value

The two mating systems also caused hard to explain differences
in EBV averages for calves, Sires and dams. The average EBV for
calves are presented in Table 5. Part of the difference in EBV average
for the two mating systems may be due to the extra phenotypic variance
in YW that PAM has over RM. Sire and dam EBV averages followed the
pattern set by the calves, with the difference between the two mating
systems increasing over time. At CClO PAM sires and dams were 25 and
14% higher, reSpectively, in their EBV average. These differences in
sire and dam EBV averages were not caused by differences in number of
calves produced by each sire and dam. At CClO the number of offSpring
per sire averaged 20.1 and 20.6 under RM and PAM, respectively, while
cows averaged 4.2 and 4.1 calves under RM and PAM, respectively.

Even though mating system did effect EBV averages no harmful
effect on its use seemed to have occurred. Thus it is safe to assume

EBV will be useful regardless of mating system.

78

 

 

_.Noo Noo.o o.ooo o.oNo ooo.N o.ooo o.ooo ooN. N.ooo ooo
N.ooo o»o.N o.o»o o.»oo ooN.N N.ooo _.ooo 3oo.N o.oNo 3» oocoooco
o.ooo »oo.N »._»o ».3oo N»o.N N.3oo N.ooo moo.N _.oN¢ »oo
oooo.v oooo.v mooo.v oooo.v mooo.v oooo.v »_o. oooo.v oooo.v .o»o oo_ooo_om
o.Noo o_o.o N.oom o.o3o ooo.N o.ooo o.ooo »_N.N o._No 3oo
».3oo Noo.o o.ooo o.o3o moo.N _.ooo o.ooo ooo.N _.NN3 3oo oo>oo
N.ooo N»o.N o.Noo o.oNo ooN.N N.NNo o.ooo 33o.N o.ooo 3oN moo3_a3
oooo.v oooo.v oooo.v oooo.v mooo.v oooo.v mooo.v ooo. Noo. .oom ooooooo
N.oNo oNo.N N.ooo o.omo Noo.N ».moo o.ooo Noo.N o.ooo o\o
o.oNo moo.N o.»om o.ooo ooo.N o.oo3 o.ooo NoN.N o._No o\N ooooooooooo
o.»No ooo.N o.ooo o.ooo oNo.N o.oo3 o.ooo ooo.N o.ooo o\o
ooo. N»N. Noo. oNo. Noo. o_3. oNN. ooo. Noo. .oom »»oo
_.NNo» om».o P.oNo N._No moo.N 3.ooo N.o_o ooN.N o.333 3oo
o.oNo N»o.o o.3o3 o.ooo Noo.N o.ooo N.Noo N_N.N N.oo3 3o3 »o»»ooooooo3
o.»No ooN.N o.oN3 o.ooo ooo.N _.oNo o.»oo o_o.N _.oo3 3oN
mooo.v mooo.v mooo.v mooo.v mooo.v mooo.v oooo.v oooo.v mooo.v .o_o 3»
o.»oo Noo.N o.Noo N.oNo ooN.N o.oo3 o.ooo NNo.N o.ooo »o»oaaoooo oo_oma»
o.ooo NNo.o o.moo o.N3o ooo.N o.ooo 3.3oo N»N.N _.oNo o_3oo3
mooo.v oooo.v mooo.v oooo.v mooo.v oooo.v mooo.v mooo.v mooo.v .o»m oa_o
3» oo< 33 3» oo< 33 3» oo< 33, 3goooooo mocoom
o» oozo e_oo N ooco a_oo o ooao o»oo

 

.3> mz< wo< .3: men <h<o

wthmmh mmmm zH moz<Hm<> mo mmumaom zH<z
mom mz<mz no hmug oz< muz<Hm<> mo m~m>4<z<

.e m4m<h

79

 

 

 

No.» oo.» No.» oo< 33.» 3o. 3o.» oo<
oo.» o».» NN.N 3» a_ooooco oo.N oo.N oo.N 3» o»aoo»co

No.N No.N oo.» »oo No.N No.N oo.N »oo
oooo.v mooo.v ooo. .ooo oooooo»oo »oo. mooo.v N»o. .o»m oo_ooo»om

oo.» oo.» No.» 3oo NN.N oo.N o». 3oo
»».N oo.» oo.N 3oo oo»om oo.N oo.N oo.N 3oo oo>om.
No.» mo.» 33.» 3oN mcooooI oo.N m».N No.» 3oN moot»o3
ooo. o_o. 33o. .ooo ooaocoo moo. moo. o»o. .o_m oooocoo

No.N oo.N »o.o o3o oo.o oo.o No.o o\o
oo.N _o._ oo.N «\N oo»ooo»osoo »o.o No.N oo.N o\N oooooo»osoo

N3. om. No. o\o »». mm. »o. o\o
oooo.v oooo.V oooo.v .o»m »»3o oooo.v oooo.v mooo.v .o»m _»3o

o». NN. oo. 3oo oo. oo. »o.» Noo
»o.N o».N »o.N 3o3 »N»»Pooooco3 NN.o o». oo.N 3oo »u»»»ooo»3¢3

3o.N oo.N o3.N 3oN NN.o oo.N »o.N 3oN
oooo.V oooo.v oooo.v .o»m 3» oooo.v mooo.v mooo.v .oom 3»
o3.» oo.» oo.» »o_ocaEEoo oo»omo» oo.N oo.N oN.N 333 soom»m

No.» No.» No.» o_oo_3 No.» No.» mo.» 33
oNo. Noo. moo. .o_m atom mooo.v oooo.v omo. .o_m oooooz

o»oo Noo 3oo 33oooooo ooaoom o»oo:: ; Noov ooo. »3oooooo. ooaoom
.<»<o oz»»mo» meow oz< <»<o zo»m»m
oZHooz 33 3mm moo ooz<oo<» 3o moo3oom 3333
moo mzeoz 3o »moo oz< Noz<oo<» 3o mom»o<z< .m Noo3»

80

Sire Testing Procedures

Effect gg_Yearling Weight

It is readily evident from data presented in Table 4 that progeny
testing in a commercial test herd, as done in this investigation,
proved to be unsuccessful at keeping pace in improvement of any trait
when campared to progeny testing within the purebred herd. At CClO
the populations utilizing the commercial test herd were 97.3% as heavy
as the other sire testing procedure. The same comparison at CC4 was
98.4%, pointing out that the difference between the sire testing
procedures increases more rapidly in the first four calf crops because
the second calf crOp cannot be sired by selected sires. However, the
difference between the two procedures continues to increase over all
calf crOps because of the increased generation interval testing in a

commercial test herd brings about.

Effects gg_Weaning Weight and Average Daily Gain

 

The component parts of yearling weight reacted similarly to Sire
testing procedures., When the pOpulations utilizing the commercial
test herd were compared to sire testing within the purebred herd, it
was found in CClO that the former was 97.5 and 97.0% as heavy for WW
and ADG, reSpectively. In calf crops 4, 7 and 10 sires in the
commercial progeny testing runs were Significantly (P<.Ol) lower in
their direct genic values for WW, maternal ability (MA) and ADG than

were sires in the purebred herd testing procedure. When the dams in

81

the two sire testing procedures were compared the differences were not
as striking, but the pOpulations utilizing the commercial test herd
were significantly (P<.Ol) lower in dams' direct genic value for WW,
MA and ADG.

In general, the progeny testing of bulls within a commercial test
herd, as done in this investigation, would be fruitless when trying to
accomplish greater improvement in the performance traits looked at in
this study. Modifications to the system, such as: .testing fewer
bulls (i.e. increasing the first stage selection intensity), using a
portion of the selected untested bulls in the purebred herd before
their test is completed, or combining progeny test with individual

performance, may give more advantageous results.
Yearling Weight Heritability

Effect gg_Performance Traits

 

In the three performance traits analyzed yearling weight
heritability (YWH) usually accounted for the major portion of the total
variance in each trait. As indicated in Tables 3 and 4, the means of
different YWH levels always appeared to be significantly different from
one another. This also appeared to be true for the genic values of the
traits presented in Tables 3 and 4. The only exception to the above
statement was genic maternal ability; in that, there was no significant
difference in the amount of improvement made in genic MA under the

three YWH levels.

82

In general WW, ADG and YW reSponded in a linear fashion to
increases in YWH. However, WW tended to deviate from linearity due
to a larger increase in its heritability when YWH went from 40 to
60%. As shown in Table 2, WW heritability increased from 19.70 to
39.25% when YWH increased from 40 to 60%, whereas, only a 10% increase
in WW heritability was used between 20 and 40% YWH. Even though
increases in heritability for ADG were about equal across the three
YWH levels, the largest increase in amount of improvement made in ADG
came between 20 and 40% YWH.

Effect gg_Estimated Breeding Value and Number gt_0ffsgring
Per Parent

 

 

AS explained in the Review of Literature, EBV are expressed as
estimated genetic differences from the herd average at one particular
time. Thus, the EBV do not give any relative genetic rankings among
herds. However, for within herd comparisons, EBV can be used to
compare individuals which do not greatly differ in age. Many sources
which could significantly effect EBV as a predictor of breeding value
have been imposed, but only the ones discussed in sections following
have caused changes in EBV averages.

The highest level of YWH caused a highly significant depression
in average EBV for calves. As Table 5 indicates, the highest YWH
level is significantly lower than either of the lower levels of YWH
for calves' average EBV. The 20% level of YWH consistently had the

highest EBV average. Sires and dams were just the opposite in the

83

ranking of EBV averages by YWH level. In CClO of the mating system
analysis the average sire EBV for 20, 40 and 60% YWH were 17.3, 25.6
and 38.2, respectively, while average cow EBV for the YWH levels in
the same analysis were 14.7, 24.5 and 34.8, respectively. The
differences in EBV averages for sires and dams by YWH level in the
sire testing analysis were almost the same as those previously
mentioned and, therefore, will not be listed.

Level of YWH significantly (P5.Ol) affected the number of calves
had by each sire and dam in CC7 and CClO. In the mating system
analysis,the number of calves per sire for YWH levels of 20, 40 and
60% in CClO were 21.3, 20.4 and 19.3, respectively. The number of
calves per cow for the same YWH levels in CClO were 4.3, 4.1 and 4.1,
respectively. Although significance (P<.Ol) did exist in number of
calves per sire or dam, it is doubted that these differences contri-
buted to the differences seen in EBV averages for the three levels of
YWH. One can speculate that an apparent contribution to the
differences in average EBV for sires and dams as heritability increases
is the higher correlation which exists between individuals and their
relatives. But why calves differed from this is not known and cannot

be explained.

Bull Combination

Effect gr; Performance Traits

 

The different combinations of untested bulls and progeny tested
l>ulls (UT/PT) had no significant effect on the amount of genetic

improvement made in YW over time. Even though Table 3 presented' a

significant phenotypic difference for YW in CClO, a significant
difference in YW genetic ability did not exist in that calf crop.
The 2/4 bull combination appeared to bring about more phenotypic YW
improvement, however, when the YW genetic abilities for each
combination were investigated a nonsignificant difference was found.
Although no phenotypic significant difference existed for WW, the
2/4 bull combination became significantly higher than either of the
other bull combinations for direct genic value for WW at CClO.
Table 3 and 4 further show that the 2/4 bull combination gave the
highest averages for the three calf crops and three traits analyzed,
with the exception of ADG in CC7 of the mating system analysis.

When the phenotypic variances of WW, ADG and YW were analyzed
no significant differences were found between the bull combinations.
In fact, the variances for all bull combinations were within one-

fourth of one standard deviation for all three traits.

Effect gg_Estimated Breedigg Value

 

As might be theorized, bull combinations significantly affected
EBV average for the calf crOp. Because selection of older bulls was
based on progeny record, an automatic positive data item was
incorporated into the EBV calculation of future offspring of the
selected sires. Thus, most of the differences in EBV averages for
bull combinations seen in Table 5 are likely due to the differences
in PHS data incorporated into the EBV calculation for YW. The number
Of calves sired by each bull for the bull combinations 0/6, 2/4 and 3/3

in the mating system analysis were 13.5, 20.9 and 24.8, reSpectively.

85

Percent Heifer Calves Saved

Effect gg_Yearling Weight

 

Percent heifers saved caused more significant differences in
the performance traits than any other entity investigated which can be
controlled by a breeder. As percent heifer calves saved (PHCS)
increased more improvement in YW was realized. Tables 3 and 4 show
that most of the phenotypic improvement in YW caused by PHCS was when
the replacement rate increased from 20 to 50%. Genetic ability for
YW reacted exactly as phenotypic expression did to PHCS, with signifi-
cance (P<.Ol) existing in the same ranking for the three calf crOps
analyzed.

Different levels of PHCS also caused significant (P<.Ol)
differences to occur in phenotypic and genic variances in CC7 and
CClO. In CClO, the 20% level of PHCS had 6 and 15% more phenotypic
variance than the 50 and 80% levels, respectively. The same compari-
son for genic variance was 16 and 28%. This result could be explained
by the fact that only the tOp 20% of the heifer calves are included in
the mating pOpulation, thus creating a subunit of females in the herd
which phenotypically average 1.4 standard deviations above the herd
mean. With a 20% replacement rate 8.1 heifer calves will be put into
the breeding population each year. With a 5% death loss in cows this
allows the culling of only 3.6 cows per year, which is only 4.2% of the
85.5 cows subject to culling. Thus the cows after culling average less
than .1 of a standard deviation above the herd mean, therefore, explain-
ing why the Substantial differences in variances occurred with the

different replacement rates.

86

Effect gg_Weaning_Weight

 

The component parts of YW, WW and ADG, were affected in the same
way by the different levels of PHCS. Switching PHCS from 20 to 50%
accounted for most of the improvement in WW caused by increasing the
percent heifers saved. In the three calf crOps analyzed, the 80%
saving rate was. at the most, 5.7 lbs. heavier in ww than the
50% rate and usually less than 2.0 lbs. heavier. Direct genic value
for WW was the highest for the herds incorporating 50 and 80% sav-
ing rates, but these two were not significantly different from one
another. Level of PHCS did not significantly effect the amount of
phenotypic or genic variance in WW.

PHCS proved to have a major effect on improvement of genic
maternal ability. The data in Table 6 Show that increased heifer
saving rates improved MA. However, a leveling Off occurred in
the mating system analysis and a nonsignificant depression in the sire
testing analysis occurred when PHCS changed from 50 to 80%. From the
slight improvement gained, it appears the added expense of developing,
breeding and calving out an additional 30% of the heifers could not
be made economically feasible. This fact will be even more evident

when the unadjusted weaning weights for bull calves are compared.

Effect gg_Average Daily Gain

 

The effects of PHCS on ADG were similar to what happened for WW.
There was no significant difference between 50 and 80% heifer sav-

ing rates for phenotypic expression or genic value. Indeed, the

87

TABLE 6. CALF AVERAGE FOR GENIC MATERNAL VALUES
IN MATING SYSTEM DATA AND SIRE TESTING DATA.

 

Mating System Analysis Sire Testing Analysis
Source Category CC4 CC7 CClO CC4 CC7 CClO

 

Percent Sig. .052 .007 .031 .083 .023 .046
Heifers 20% 2.3 4.5 8.2 1.5 4.2 6.9
Saved 50% 3.8 8.9 12.5 4.0 8.2 11.4
80% 5.2 10.0 13.3 4.0 7.6 10.1
Selection Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005
EBV 5.0 10.8 14.4 4.7 9.0 11.9
Criteria YW 7.8 12.7 19.4 6.1 12.0 16.3
ADG -1.5 -.1 .1 -1.4 -1.1 .2

 

TABLE 7. BULL CALF AVERAGE FOR UNADJUSTED WEANING WEIGHT
IN MATING SYSTEM DATA AND SIRE TESTING DATA.

 

Mating System Analysis Sire Testing Analysis
Source Category CC4 CC7 CClO CC4 A CC7 CClO

 

Percent Sig. <.0005 .002 <.0005 <.0005 <.0005 .003

Heifers 20% 455.4 467.2 468.6 456.8 467.8 467.1
Saved 50% 447.8 465.7 481.7 448.2 462.5 477.4

80% 442.0 458.2 477.7 438.5 454.9 470.0

Selection Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005
EBV 454.6 475.4 490.4 452.7 471.9 484.8

Criteria YW 455.3 476.4 492.5 454.1 473.5 487.0
ADG 435.4 439.4. 445.1 436.8 439.7 442.7

 

88

 

 

 

 

 

TABLE 8. BULL CALF AVERAGE FOR UNADJUSTED YEARLING WEIGHT
IN MATING SYSTEM DATA AND SIRE TESTING DATA.

Mating System Analysis Sire Testing Analysis

Source Category CC4 CC7 CClO CC4 CC7 CClO

Percent Sig. .074 .001 <.0005 .008 .132 <.0005

Heifers 20% 886.0 917.8 938.2 883.0 915.2 932.5

Saved 50% 886.0 932.3 974.8 879.6 918.6 957.8

80% 880.0 924.9 975.5 872.9 911.7 952.7

Selection Sig. .006 <.0005 .012 .037 <.0005 <.0005

EBV 888.6 933.2 967.6 881.9 922.4 953.8

Criteria YW 884.5 926.2 965.7 879.8 918.0 953.3

ADG 878.8 915.6 955.2 873.7 905.0 939.9

TABLE 9. MEANS AND SIGNIFICANCE VALUES 0F EBV AND
YW BREEDING VALUES FOR SIRES AND DAMS IN
MATING SYSTEM AND SIRE TESTING DATA.
Mating System Analysis Sire Testing Analysis
Selection

Source Criteria CC4 CC7 CClO CC4 CC7 CClO

Mean of Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005

Sires' EBV 27.9 29.3 30.2 25.8 25.2 26.0

EBV YW 26.5 26.4 29.3 25.0 22.0 23.1

ADG 19.1 18.6 21.5 17.4 15.6 17.2

Mean vw Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005

Breeding EBV 947.4 986.5 1025.5 924.8 968.1 1005.4

Value for YW 930.7 978.4 1028.0 923.2 965.9 1009.1

Sires ADG 923.7 967.0 1011.6 913.9 949.0 990.1

Mean of Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005

Dams' EBV 19.2 26.7 31.0 19.1 26.1 30.1

EBV YW 16.1 22.3 25.2 15.7 21.6 23.9

ADG 12.0 15.6 17.9 11.2 14.7 16.4

Mean yw Sig. <.0005 <.0005 <.0005 <.0005 <.0005 <.0005

Breeding EBV 887.0 927.5 965.4 884.1 919.4 953.2

Value for YW 884.1 924.4 966.7 881.5 918.8 955.3
Dams ADG 878-3 91395 954:? 876-1 9073. 9418,

 

89

additional heifers saved over 50% in this study gave no beneficial
result in the way of increased improvement in ADG. However, both
50 and 80% saving rates were Significantly higher in phenotypic
and genic value averages than the 20% saving rate.

When the phenotypic and genic variances for ADG were analyzed
the results were parallel to those obtained for YW. In the mating
system analysis data at CClO, the 20% 1 saving rate had 19% more
phenotypic variance and 34% more genic variance than 50 and 80% PHCS.
In CC7 the differences were also significant, but not as large. There

were no significant differences in CC4 for these two variances.

Effect gg_Unadjusted Weaning Weight and Yearling Weight

 

Both analyses in Table 7 show that 20% heifer saving rate
was superior in its average unadjusted WW for bull calves in CC4 and
CC7 when compared to either 50 or 80% saving rates. However,
by CClO enough additional improvement in direct genic value for WW and
MA occurred in the 50 and 80% saving rate pOpulations SO that
their unadjusted WW averages were superior to the 20% replacement rate.
Table 8, which gives the means for unadjusted YW, shows a different
reaction to heifer saving rates. Because ADG is not adjusted for
age of dam effect, the additional improvement which was made in ADG
with the two higher PHCS levels made up for the unadjusted WW
depression which occurred in the earlier calf crops. AS presented in
Table 8, the 80% PHCS level was inferior to 50% on all but one occasion

for bull calves.

9O

Heifer calves responded exactly as bull calves did to the three
levels of PHCS for both unadjusted WW and YW. But the unadjusted WW
for heifer calves was on the average 10% less than bull calves due to
adjustment to a common sex basis. Naturally, this 10% difference at

weaning time was carried through to the unadjusted YW.

Effect gg_Generation Interval

 

Even though PHCS had no significant effect on average calf EBV
in the three calf crOpS analyzed, it did effect the generation
interval. As PHCS decreased, the generation interval increased due to
a greater number Of years the cows were left in production. At CClO
in the mating system analysis, 20, 50 and 80% heifer saving rates
averaged 6.1, 3.7 and 2.8 calves per cow, respectively, and in the
sire testing analysis the same respective averages were 6.1, 3.6 and
2.8. The overall mean number of offSpring per cow for both analyses
and the three PHCS levels was 4.2. From the above figures, one can
easily calculate an estimated turnover rate for the females. It is
equal to the average number of offSpring per cow, plus 10% of that
value to allow for calf death loss, plus 1.0, which corrects the cows'
age to what will happen under normal production schemes, as females
in the simulated pOpulations reproduce at one year of age. Therefore,
from this method the turnover rates in the mating system data were
7.6, 5.1 and 4.1 years, reSpectively, for 20, 50 and 80% PHCS.

Even though selection differentials for heifers in the lowest
saving rate would be much larger than the other two rates, the

selection differential on the mature cows would be extremely small.

91

Because of this, plus the low turnover rate, the depression in adjusted
yearling weights with the low PHCS Should be of no surprise. Indeed at
the low heifer saving rate one is not taking advantage of the
genetic reach one made when the selection of their sires was done.

From this data one would not recommend a low heifer saving rate,

for it would be wasteful of the genetic trend brought about by the
intense selection for the Sires. As stated previously, putting 80%

of the heifers back into the breeding herd would not be economically
feasible because of the additional expenses and small amount Of
additional improvement made with this larger array of heifers. It is
this investigator's estimate that somewhere between 50 and 65

percent heifer calves saved will maximize genetic improvement without an
intolerable number of actual pounds lost due to the age of dam

adjustment.
Selection Criteria

Effect on Yearling Weight

The selection methods utilized in this investigation highlighted
what might be predicted and also brought out problems which may need
to be resolved with future research. Tables 3 and 4 Show that in this
study the most improvement in YW came when selection was made on YW
itself or the estimate of its breeding value, EBV. EBV selection
consistently gave greater improvement in YW through CC7, but by 0010
YW selection caused heavier weights to occur, although EBV and.YW.
selection never appeared to be significantly different from one another

in the three calf crOps analyzed. Table 9 presents why these two

92

selection methods may have switched in their order as time progressed.
In all three calf crOps of the mating system analysis, EBV selection
had the highest sire and dam average for EBV. However, CClO data in
that analysis Show that YW selection gave parents which had superior
genetic ability for YW. In the Sire testing analysis these same
differences were even larger. Also to be noted from Table 9 is that
cows averaged as high in EBV as sires, even though the intensity of
selection was much greater for sires. This indeed points out that a
serious bias in female EBV may be occurring; in that, as years go by
their individual deviation from herd mean is not readjusted and the
records for calves produced in the early part of a cows' production
are included in the EBV calculation as deviations from means that
are below the genetic mean of the herd at the time the dams are
selected. Thus,the EBV of selected females is biased upward in herds
that are improving rapidly from one calf crOp to the next.

When ADG selection was compared to EBV and YW selection it
averaged significantly (P<.Ol) lower in YW improvement. The reason
for this is the Slow improvement in WW.

Selection criteria caused no Significant change in the amount of
phenotypic or genic YW variance. EBV selection did, however,
consistently give populations With the most phenotypic and gen1c YW

variance.

Effect on Weaning Weight
ADG selection gave slower improvement in YW than either EBV or

YW selection. The main cause of this slower improvement was the

93

extremely slow correlated improvement in direct genic value for WW.
Tables 3 and 4 show that by CC4 YW and EBV selected populations
averaged approximately 20 lbs. higher than ADG selected populations

in their WW performance. Of course, the main reason why this is occur-
ring is that WW is a component part of YW. When selection is practiced
on either YW or EBV a great deal more selection pressure is placed on
direct genic value for WW than when selection is by ADG. Also
suffering heavily when ADG selection is employed is MA. Since no
correlation existed between ADG and MA, the latter neither increased
nor decreased as ADG increased (Table 6). Due to the large deficit
these two aspects summed up to, selection by ADG did not result in
substantial enough improvement in ADG over YW and EBV selection to
make it a competitive way to optimize improvement in YW.

The greatest amount of correlated improvement in genic value for
WW was when YW selection was practiced (Table 3 and 4). However,

YW and EBV selection appeared not to be significantly different.
Selection criteria affected genic value fOr MA the same way it did
phenotypic and genic values for WW. YW selection, as seen in Table 6,
gave the largest improvement in genic value for MA, while EBV
selection improved it at a Slower rate and ADG selection caused no
change.

Selection criteria brought about no significant changes for
either phenotypic or genic variances in WW. In CC4 and CC7,YW select-
ion was lower in its phenotypic and genic variance than the other two
selection criteria, but in CClO, YW selection was intermediate in its

variances.

94

tttggt_gg_Average Daily Gain

As should be expected, ADG selected populations had the greatest
amount of improvement in ADG. In the mating system analysis (Table 3),
ADG selected populations at CClO averaged .22 lbs. per day higher
than EBV and YW selected pOpulations, while in the sire testing
analysis (Table 4), ADG selected populations averaged .16 lbs. per day
higher than those selected on EBV and YW. Without doubt, if ADG is
the single, utmost important trait to improve in the beef industry,
then selection by ADG or by an estimated breeding value of it is the
route one should follow to the highest degree. However, as previously
mentioned,ADG selection will not bring about Optimum improvement in
yearling weight because of its downfall in increasing performance up
to weaning time. EBV and YW selection did not differ significantly
in their improvement rates for ADG.

As with the previous performance traits, selection criteria did

not:affect the amount of phenotypic or genic variance in ADG.

Effect gg_Unadjusted Weaning and Yearling Weights

 

Unadjusted WW for bull and heifer calves reSponded to selection
criteria exactly as adjusted WW did. Even though YW and EBV selection
appear not to be significantly different from one another in unadjusted
WW, Table 7 Shows that YW selected populations were consistently
higher than EBV selected populations. YW and EBV selected populations
were significantly (P<.Ol) superior in unadjusted WW performance when
compared to ADG selected populations. This last statement also held
true for unadjusted YW, thus making YW and EBV selection the most

desirable for improving actual weight at one year of age.

95

Effect gg_Estimated Breeding Value

 

Selection criteria had a major impact on calf crop averages for
EBV. As Table 5 presents, EBV and YW selection in both the mating
system and sire testing analyses gave consistently higher calf crop
averages for EBV than did ADG selection. The way YW and EBV selected
populations ranked in EBV average differed for the two analyses. In
the mating system analysis,YW selected populations were higher in
EBV average than those using EBV selection, whereas in CC7 and CClO
of the sire testing analysis, EBV selected populations were higher
than those selected by YW methods. When EBV averages for sires and
dams were analyzed the same pattern of events that were described
above occurred. It is likely a lower correlation between ADG and
estimated breeding value for YW existed, thus causing ADG selection
to average lower in calf crop EBV. At this time, the relevance of
this to genetic improvement is indeed questionable.

Selection criteria had a highly Significant (P<.Ol) effect on the
number of offspring per sire and dam. The number of calves per sire
and dam in both analyses averaged higher for EBV selection than for
either ADG or YW selection. The average number of calves per Sire for
EBV, YW and ADG selection in the mating system analysis were 22.0,
19.4, and 19.6, respectively, while in the sire testing analysis the
same respective averages were 22.5, 18.1 and 18.5. The two analyses
averaged the same for number of calves per cow,with EBV, YW and ADG
selection averaging 4.4, 4.1 and 4.0, respectively. Indeed it appears

that selection by EBV will tend to keep a cow in the breeding herd for

96

close to a half year more than YW or ADG selection. This further
emphasizes the point made earlier about the bias in EBV calculation
for older cows; in that, the cows' individual deviation from the
herd mean and the performance of offspring in past calf crops for
YW tends to bias the EBV upward, thus making the retention period

for females in the breeding herd longer.

General

In short, it appears that either EBV or YW selection will
maximize improvement in most of the performance traits. Although
only small differences exist between EBV and YW selection, most of
these lie in favor of YW selection, thus making it slightly more
favorable in maximization of performance improvement. If modifica-
tiOns to individual deviation from herd mean could be employed,
EBV selection may Show up as a more desirable tool for selection.
However, when the deviation from the herd mean for an old animal
is considered to be equal to an individual deviation from the
current herd mean,serious biases result and consequently inaccurate

culling of animals takes place.

Interactions Between Main Sources gt_Variance

 

Mating System and Yearling Weight Heritability

When the analysis of the mating system data was done this inter-
action appeared to affect both ADG and EBV. For ADG, mating system
did not interact with YW heritability (YWH) in CC4, but the level
of Significance was .037 for CC7 and .025 for CClO. The same

97

interaction was significant in CC7 and CClO for genic value for ADG.
Table 10 shows that the interaction in ADG occurs because PAM
responds better to the highest heritability level for YW. For YWH
levels 20 and 40%, ADG was equal in both mating systems, however,
with 60% YWH, ADG averaged .07 lbs. per day higher in PAM than RM.
This interaction was also noted in CClO of the analysis Of genic
value for ADG variance. At the 60% level of YWH, PAM had 28% more
genic variance than RM at the same heritability, whereas at the other
two levels RM and PAM were within 10% of one another.

In CC4 and CClO, mating system Significantly (P<.Ol) interacted
with YWH for calf crop EBV average. Table 11 shows that when YWH
increased from 40 to 60% a much greater reduction in EBV average
occurred with the RM populations than with PAM populations. To
properly explain why this situation developed, an investigation of
the component parts used in calculating EBV would be required. Because
these data items were not collected no solution will be offerred, but
the fact that the same significant (P<.05) interaction existed in
the three calf crops analyzed for sire and dam EBV averages will be
added to strengthen the evidence that there was an interaction between
mating system and YWH. The relevance of this to performance trait
improvement and the ramifications it may have are unknown at this time.
However, it is an occurrance which breeders should be aware of, but
one which will have no major impact on improvement programs. The
EBV is not an end to a mean, but an aid, which, when properly put
into perspective, can be extremely helpful in the selection of superior

genetic stock within a population. It is pertinent for breeders to

98

TABLE 10. ADG MEANS FOR MATING SYSTEM AND
YW HERITABILITY SUBCLASSES.

 

 

Marginal

20% 40% 60% Means

CC4 2.63 2.74 2.78 2.72

RM CC7 2.72 2.90 2.99 2.87
CClO 2.78 3.07 3.22 3.02

CC4 2.63 2.74 2.82 2.73

PAM CC7 2.71 2.92 3.06 2.90
CClO 2.80 3.08 3.31 3.06

Marginal CC4 2.63 2.74 2.80 2.72
Means CC7 2.71 2.91 3.03 2.88
CClO 2.79 3.07 3.26 3.04

 

TABLE 11. EBV MEANS FOR MATING SYSTEM AND YW
HERITABILITY SUBCLASSES.

 

 

Marginal

20% 40% 60% Means

CC4 2.66 2.34 .78 1.93

RM CC7 2.70 2.42 .49 1.87
CClO 2.86 3.00 -.09 1.92

CC4 2.37 2.54 1.85 2.25

PAM CC7 2.97 3.15 1.38 2.50
CClO 3.58 3.54 1.87 2.99

Marginal CC4 2.51 2.44 1.31 2.09
Means CC7 2.84 2.78 .94 2.18
CClO 3.22 3.27 .89 2.46

 

99

remember that different selection intensities, turnover rates and
YWH levels can cause major changes to occur in population means for

EBV.
Sire Testing and Yearling Weight Heritability

This interaction occurred in CClO for both A06 and YW. The
Significance level for both performance traits was .016. Table 12
shows why in CClO this Significant interaction occurred in ADG and
YW. The interaction was caused by a greater increase in both ADG
and YW as YWH increased for the case of within purebred herd sire
testing than for sire testing in a commercial test herd. At the low
level of heritability, progeny testing within the purebred herd
proved to be below its expected value based on the marginal means.
When the YWH was 60%, within purebred herd sire testing exceeded its
expectation and sire testing within a commercial test herd fell short
of its expectation. Indeed these results back up what has been
theorized; that is, as heritability of the trait increases, there
is less advantage in progeny testing of males before their inception

into the purebred herds.
Sire Testing and Percent Heifer Calves Saved

The interaction of these two independent variables occurred in
YW and unadjusted YW. For the former trait, the interaction was
significant (P§.Ol) in CClO, while it was significant (P<.O5) in both
CC7 and 6010 for unadjusted YW. The interaction was the result of

100

TABLE 12. ADG AND YW MEANS IN CALF CROP 10 FOR SIRE
TESTING AND YW HERITABILITY SUBCLASSES.

 

 

20% 40% 60% Marginal Means
ADG YW ADG YW ADG , YW ADG YW
Within 2.78 928.6 3.07 991.1 3.22 1045.5 3.02 988.4

Commercial 2.74 914.5 2.97 962.4 3.08 1008.6 2.93 961.8

Marginal 2.76 921.5 3.02 976.8 3.15 1027.1 2.98 975.1
Means

 

TABLE 13. YW AND UNADJUSTED YW MEANS IN CALF CROP 10 FOR
SIRE TESTING AND PERCENT HEIFER CALVES SAVED

 

 

SUBCLASSES.
20% 50% 80% Marginal Means
YW U.YW, YW_' U-YW, YW_ U.YW.' YW , , U.YW
Within 960.0 939.1 1000.1 973.5 1005.1 970.2 988.4 960.9

Commercial 947.4 925.9 968.1 942.1 970.0 935.3 .961.8 934.4

Marginal 953.7 932.5 984.1 957.8 987.5 952.7 975.1 947.6
Means .

 

101

within purebred herd sire testing responding to increased heifer
saving rates more than sire testing within a commercial test

herd (Table 13). The difference between sire testing within the
purebred herd and in the commercial test herd was less at 20% PHCS
than at 50 or 80% PHCS.

The two interactions reported on sire testing, indeed, reinforce
the thinking that in order for extensive progeny testing of future
sires to be warranted, heritability of the trait must be low and
heifer saving rates limited because of costs or facilities
necessary for maintaining the replacements before their inception into

the productive herd.
Yearling Weight Heritability and Selection Criteria

Interaction between these independent variables occurred
consistently in WW, ADG.and unadjusted WW in both analyses and in all
three calf crops analyzed. This interaction also occurred for MA in
CC7 and C010 of the mating system analysis. Mating system analysis
data are presented in Table 14 to Show that ADG selected populations
did not respond as much to YWH increases as EBV and YW selected
populations reSponded. CC4 and CC7 data were the same as CClO, only
the degree of difference was not as great. As should be expected,
unadjusted WW for bull and heifer calves responded in the exact
manner adjusted WW did.

This interaction was highly significant (P<.Ol) for ADG in both

mating system and Sire testing analyses for the three calf crOps

102

TABLE 14. WW AND MA MEANS IN CALF CROP 10 FOR YW
HERITABILITY AND SELECTION CRITERIA FOR
MATING SYSTEM DATA.

 

EBV YW ADG Marginal Means
WW MA WW MA WW MA WW MA

 

20% 488.0 11.9 497.7 22.5 461.0 -2.3 482.2 10.7
40% 514.5 20.9 516.8 18.5 470.2 -.4 500.5 13.0
60% 547.7 10.5 549.8 17.2 481.9 3.2 526.5 10.3

Marginal 516.7 14.4 521.4 19.4 471.0 .1 503.1 11.3
Means

 

TABLE 15. ADG MEANS FOR THE INTERACTION OF YW
HERITABILITY WITH SELECTION CRITERIA
FOR MATING SYSTEM DATA.

 

 

Marginal
EBV YW ADG Means
CC4 2.64 2.60 2.65 2.63
20% CC7 2.71 2.68 2.74 2.71
CClO 2.76 2.77 2.85 2.79
CC4 2.71 2.71 2.80 2.74
40% CC7 2.87 2.83 3.02 2.91
CClO 2.99 2.99 3.24 3.07
CC4 2.80 2.73 2.88 2.80
60% CC7 3.00 2.93 3.16 3.03
CClO 3.18 3.14 3.47 3.26
Marginal CC4 2.72 2.68 2.78 2.72
Means CC7 2.86 2.81 2.97 2.88
CClO 2.98 2.96 3.19 3.04

 

103

investigated. Table 15, with data from the mating system analysis,
shows ADG selection gave more improvement in ADG than either EBV
or YW selection as YWH increased. It is of interest to note that
EBV and YW selection were almost identical in the amount of improve-
ment they caused in ADG at all levels of YWH. Data from the sire
testing analysis for this interaction were a replicate of mating
system data, except the results of ADG selection were not as
drastically different from EBV and YW selection.

Yearling Weight Heritability and Percent

Heifer Calves Saved

These two independent variables interacted to a significant
degree in YW (P<.Ol) and ADG (P<.05) in the three calf crops analyzed.
However, this interaction was significant only in the mating system
analysis. Table 16 shows that linear increases in improvement
occurred across PHCS levels with YWH levels of 20 and 40%. But, with
the 60% level of YWH,a leveling off occurred in ADG and YW when PHCS
changed from 50 to 80%.

Significance of this interaction in WW was sporadic with no set
pattern occurring in the parts comprising the makeup of WW. The
interaction was significant in CC7 and €010 for direct genic value
of WW, but significant for WW phenotype only in CC7. Significance
(P<.05) for MA occurred only in CC4 and this was due to a negative
response as PHCS increased with the heritability level at 60%,
whereas with the other YWH levels MA increased as PHCS increased.

Direct genic value and phenotypic averages for WW reSponded to increases

104

 

 

 

 

 

TABLE 16. ADG AND YW MEANS FOR YW HERITABILITY AND
PERCENT HEIFER CALVES SAVED IN MATING
SYSTEM DATA.
20% 50% 80% Marginal Means
ADG YW ADG YW ADG YW ADG YW
CC4 2.62 880.6 2.63 884.1 2.64 889.6 2.63 884.8
20% CC7 2.67 895.0 2.72 909.7 2.75 921.6 2.71 908.8
CClO 2.72 907.3 2.80 932.3 2.85 946.9 2.79 928.9
CC4 2.68 893.7 2.76 910.3 2.78 921.7 2.74 908.6
40% CC7 2.81 923.7 2.94 953.5 2.96 965.7 2.91 947.7
CClO 2.94 956.6 3.10 999.4 3.17 1019.7 3.07 991.9
CC4 2.78 916.3 2.83 936.9 2.80 927.1 2.80 926.8
60% CC7 2.96 963.0 3.09 1009.4 3.03 992.0 3.03 988.1
CClO 3.15 1011.4 3.33 1069.1 3.31 1065.5 3.26 1048.7
Marginal CC4 2.69 896.9 2.74 910.4 2.74 912.9 2.72 906.7
Means CC7 2.81 927.2 2.92 957.6 2.92 959.7 2.88 948.2
’CC1O 2.94 958.5 3.08 1000.3 3.11 1010.7 3.04 989.8
TABLE 17. ADG MEANS FOR YW HERITABILITY AND BULL
COMBINATIONS IN MATING SYSTEM DATA.
a Marginal
0/6 2/4 3/3 Means
CC4 2.64 2.61 2.63 2.63
20% CC7 2.69 2.70 2.75 2.71
CClO 2.77 2.78 2.83 2.79
CC4 2.73 2.74 2.15 2.74
40% CC7 2.90 2.90 2.92 2.91
CClO 3.05 3.10 3.06 3.07
CC4 2.80 2.84 2.77 2.80
60% CC7 3.03 3.05 3.00 3.03
CClO 3.27 3.29 3.23 3.26
Marginal CC4 2.72 2.73 2.72 2.72
Means CC7 2.87 2.88 2.89 2.88
0010 3.03 3.06 3.04 3.04

 

aFirst number indicates the number of progeny

tested sires

utilized and the second number indicates the number of untested bulls.

105

in YWH and PHCS in the same manner. For YWH levels 20 and 40%, WW
increased as PHCS increased, however, at 60% YWH, WW increased when
PHCS increased from 20 to 50% and then leveled off.

The interaction was the same in all three calf crops for ADG.
Increases in.ADG phenotype and genic value were realized as PHCS
increased with the lower two YWH levels. However, when YW was 60%

heritable, the 50% heifer saving rate was superior to the 80%

rate as presented in Table 16.
Yearling Weight Heritability and Bull Combination

The only trait influenced by this interaction was ADG in the
mating system analysis. The significance of this interaction for
ADG phenotype in CC4, CC7 and CClO was .018, .045 and .069, respective-
ly. For genic value the significance levels for the same three calf
crops were .012, .068 and .049, respectively. Table 17 shows that
at low YWH more improvement in ADG is made when more progeny tested
bulls are incorporated into the breeding system. However, at the
highest YWH,the use of fewer progeny tested bulls enhanced the rate
of improvement. This supports the theory of keeping generation
interval at a minimum when heritability of the trait is high, thus
causing the accuracy of breeding value estimation based on own

performance to be at a high level.
Selection Criteria and Percent Heifer Calves Saved

Significance (P<.05) for this interaction occurred only in ADG

for the mating system data. As might be expected, this interaction

106

occurred because ADG selection responded to the levels of PHCS to a
greater degree than EBV and YW selected populations. With EBV
selection, the average improvement in ADG, when saving rate

went from 20 to 80%, was .08 lbs. per day. YW selection also had

an average improvement rate of .08 lbs. per day in ADG, but ADG
selected populations averaged .20 lbs. per day improvement in ADG

as heifer saving rate went from 20 to 80%. It was noted from

the analysis, that of the .20 lbs. per day improvement, 87% occurred

when PHCS increased from 20 to 50%.

General

In brief, many interactions between the independent variables
occurred, but in most instances there was simply a change in the
differences among responses to the major effects. Thus, major
changes in the ranking of cases in the independent variables were
seldom. The only change in ranking monitored by a significant inter-
action effect was between YWH and bull combinations. In this case,
as YWH increased a faster turnover rate in bulls became more

desirable for greater genetic gains in ADG.

CONCLUSIONS

It is pertinent that a breeder of any purebred livestock be
competent in aspects of animal breeding which he can control. This
investigation was set up to include different alternatives which
could be used by breeders of beef cattle. Indeed, at the outset of
the investigation it was known that certain options would not
maximize gentic improvement, however, inclusion of them was deemed
necessary if perspectives for the other Options were to be
established. Even though this simulation is not exactly parallel
to production schemes occurring today, it is thought that results
obtained in this study can be used as guidelines to steer a breeder
from unnecessary wrong practices, which might jeOpardize his long-
term goals of performance improvement.

When the two mating systems, random mating (RM) and positive
assortative mating (PAM), were compared, few advantages were found
in favor of PAM. PAM populations were significantly lower than RM
populations in average phenotype and direct genic value for weaning
weight (WW) in the three calf crOps analyzed. However, PAM did cause
significantly greater gains in genetic improvement for average daily
gain (ADG). But the .04 lbs. per day for 160 days was not great
enough to overcome the WW deficit which PAM populations suffered,
thus not giving PAM populations significantly superior performance
at one year of age.

When phenotypic and genic variances were compared for the two

mating systems, it was found that PAM caused Significantly more

107

108

phenotypic and genic variance to occur in ADG than RM. However,
there was no difference between the two mating systems in the amount
of variance for either WW or yearling weight (YW). These facts
point out that a purebred breeder should not spend a great deal of
time in assorting his breeding population. However, the results
still pointed out that prOper evaluation of breeding values are
necessary for Optimization of genetic improvement.

The use of a commercial test herd to progeny test performance
tested bull calves before their use in the purebred herd proved to
be fruitless in this investigation, especially when yearling weight
heritability rose above 20%. When a trait is measurable and moderate
to high in heritability, as were the traits (ADG and YW) in this
investigation, then the use of mass selection proved to give greater
genetic improvement than requiring a progeny test on bulls before
they are used. Possibly by incorporating performance tested bull
calves in the sire ranks before progeny test data is available will
make a modified progeny testing method more competitive than the
system utilized in this study.

The combinations of performance, progeny tested (within the
purebred herd) and performance, non-progeny tested bulls were not
significantly different in the amount of improvement they caused in
genetic ability for WW, ADG and YW. However, in practically all calf
crOps analyzed the combination of two performance, progeny tested and
four performance, non-progeny tested bulls was highest for phenotypic
average of WW, ADG and YW, thus giving partial evidence that the

fastest possible bull turnover rate is not the best for optimization

109

of genetic improvement. Different bull combinations caused no
changes in genic or phenotypic variance to occur.

One of the major findings in this study was the tremendous
impact heifer retention rate had on the improvement of genetic ability
for the measured performance traits. The 50 and 80% levels of percent
heifer calves saved (PHCS) caused significantly greater genetic and
phenotypic gains in all performance traits investigated when compared
to the 20% level. The levels of PHCS with the lowest selection
differentials gave more genetic and phenotypic gains than did the PHCS
level with the highest selection differential. In most cases, the
80% level was not significantly different from 50% PHCS, thus not
warranting the extra expenses involved in saving an additional 30%
of the heifers. The superiority of the 50 and 80% PHCS populations
was especially pronounced in the averages for genic maternal ability.

The weaning weights and yearling weights, not adjusted fOr the
age of dam effect, were compared for the three levels of heifer saving
rate. In the data analyzed, the 20% level was higher than 50 and 80%
levels in unadjusted WW through calf crop 7. However, by CClO the
greater improvement in direct genic value for WW and maternal ability
caused the 50 and 80% levels of PHCS to exceed the 20% level in
unadjusted WW. The 50 and 80% heifer saving rates were also superior
to 20% PHCS for unadjusted YW because of the greater genetic gains
made in ADG.

As was expected, PHCS levels had a major effect on generation
interval. The average cow age for 20, 50 and 80% PHCS were 7.6, 5.1

and 4.1 years, respectively.

110

In evaluating the effect on WW the data Show that the EBV and
YW selected populations were 45.7 and 50.4 lbs. heavier, respectively,
than ADG selected p0pu1ations in average calf WW for calf crop 10 of
the mating system data. The same figures for the sire testing data
were 42.1 and 47.4 lbs., respectively.

Selection of the basis of ADG had no significant effect in
genic maternal ability, while selection for EBV or YW significantly
increased genic maternal ability by calf crop 4. ADG selected
populations were significantly superior in ADG improvement to those
populations utilizing EBV and YW selection. Selection criterion,
however, did not significantly affect the amount of phenotypic or
genic variance in ADG.

EBV selection increased the number of years a cow stayed in
production. This was due to the fact that individual deviation from
the herd mean for a cow at the time she was a calf was used as a
constant in her EBV calculation, along with, paternal and maternal
half sib average performance and progeny performance if available.
The average genetic ability of a herd increased with each new calf
crop, but the individual deviations for the older cows were still
taken as deviations from the mean of their calf crop which was lower
than for cows in later calf crops.

Many interactions occurred between the independent variables.
The seven interactions reported on were the following: mating system
and yearling weight heritability (YWH), sire testing and YWH, sire
testing and PHCS, YWH and selection criteria, YWH and PHCS, YWH and

bull combination, and selection criteria and PHCS. All but one of

111

the significant interactions reported were of the type where the
differences among the means due to one main effect increased as the
level of another main effect changed. Thus, few changes of the rank-
ings of the means occurred. The only interaction which had a switch
in the ranking of the means was bull combination and YWH. When YWH
was low more improvement in ADG was made when more progeny tested
bulls were incorporated, whereas with the highest YWH, the use of
fewer progeny tested bulls enhanced the rate of improvement.

In general, the interactions were not alarming in their effect
on performance improvement. Five of the seven interactions involved
YWH, which, as every breeder should know, cannot be significantly
altered or controlled, thus lowering the importance of those inter-
actions involving it. The two interactions not involving YWH were

simple increased response as another source of variance changed.

SUMMARY

The data used in this study were collected from the Simulation
of approximately 295,000 animals. A biological model was used to
simulate the performance of animals based on genic effects from the
parents, Mendelian sampling and environmental effects. All selection
and mating decisions were accomplished by the use of different
computer programs.

Two different mating systems, random mating and phenotypic
positive assortative mating, were compared, along with two bull
testing procedures, testing within the purebred herd and testing
within a commercial test herd. Also included as independent variables
in the study were three levels of yearling weight heritability, 20,

40 and 60%, three different bull combinations, () progeny tested and 6
non-progeny tested, 2 progeny tested and 4 non-progeny tested, and 3
progeny tested and 3 non-progeny tested, three levels of percent heifer
calves saved, 20, 50 and 80%, and three selection criteria, EBV, YW

and ADG. When selection was on YW and ADG, the bull and heifer calves
were selected on their phenotypic expression, however, the older bulls
and cows, which had produced offspring, were selected totally on progeny
averages.

Traits investigated in the study were phenotypic and genic means
and variances for weaning weight, average daily gain and yearling
weight for calves, sires and dams. Also included in the array Of data
items summarized were EBV averages for calves, sires and dams,
plus progeny means for sires and dams and average number of offspring
per parent. Genetic and phenotypic correlations between mated parents

112

113

for average daily gain and yearling weight, plus the correlation
between EBV's of mated parents were analyzed.

The analysis of variance was done in two parts. One procedure
analyzed the two mating systems and the four factorialized sources
of variance, while the second procedure analyzed the two Sire
testing procedures and the four factorialized sources of variance.

Random mating (RM) and phenotypic positive assortative mating
(PAM) were not significantly different in the amount of improvement
they caused in YW over ten calf crops. PAM caused an insignificant
increase in phenotypic and genic variance for calf YW. The phenotypic
correlation for YW between mated individuals was .88 for PAM popula-
tions, as compared to -.06 for RM p0pu1ations. Even though the
difference in YW genetic correlation for the two mating system was
still significant (P<.Ol), it was much smaller, .16 for PAM versus
-.06 for RM.

Superior genetic and phenotypic gains in ADG were made with PAM
over RM. At calf crop 10, PAM populations averaged .04 lbs. per day
higher than RM populations in ADG. Furthermore, PAM caused signifi-
cantly greater variance in both genic values and phenotypes than RM
at calf crops 4 and 10. The average phenotypic and genetic correla-
tions between mates for the three calf crops for ADG selection in
PAM were .90 and .30, respectively, while in RM populations the same
reSpective figures were .02 and .01.

PAM was Significantly higher in its EBV average for calves, sires

and dams. At calf crop 10, PAM populations were 56% higher in their

114

calf EBV averages, however, this seemed to have no detrimental effect
on the use of EBV as a selection tool for improvement of yearling
performance .

Sire testing within a commercial test herd, as done in this
investigation, was inferior to sire testing within the purebred herd
for genetic improvement of all performance traits monitored. At calf
crOps 4 and 10, the populations utilizing the commercial test herd
were 98.4 and 97.3%. respectively, as heavy as the other sire testing
procedure at yearling time. In calf crop 10, the populations utiliz-
ing the commercial test herd were 97.5 and 97.0% as heavy fOr WW and
ADG, respectively, when compared to testing within the purebred herd.

The means for the three levels of yearling weight heritability
(YWH), for all performance traits monitored, appeared to be signifi-
cantly different from one another. Genic maternal ability was the
only trait which was not affected by YWH. WW, ADG and YW responded
in a linear fashion to increases in YWH.

The highest level of YWH caused an unexplainable significant
depression in average EBV for calves. The average EBV for sires and
dams was affected just the opposite by YWH. In calf crop 10 of the
mating system analysis, the average sire EBV for 20, 40 and 60% YWH
were 17.3, 25.6 and 38.2, respectively, while cows averaged 14.7,
24.5 and 34.8, respectively, for the same YWH levels.

The different bull combinations tested in this study caused no
significant change in the amount of genetic improvement made in YW

over time. However, data were presented to show that the combination

115

of 2 progeny, performance tested and 4 non-progeny, performance tested
bulls gave the highest averages for the three calf crops and three traits
analyzed, with the exception of ADG in calf crop 7 of the mating system
analysis. The 2/4 bull combination became significantly higher than
either of the other bull combinations for direct genic value for WW at
calf crop 10. No significant differences were found in phenotypic
variances for WW, ADG and YW for the different bull combinations
investigated.

The heifer saving rate had a positive effect on YW improvement as
percent retained increased. Fifty and 80% heifer saving rates were
significantly (P<.Ol) greater in YW improvement than the 20% retention
rate. Switching the percent heifer calves saved (PHCS) from 20 to 50%
accounted for most of the improvement in WW caused by keeping more
heifers in the breeding population. Direct genic value for WW was the
highest for the herds incorporating 50 and 80% replacement rates.
However, level of PHCS had no effect on phenotypic or genic variance in
WW.

As PHCS increased, the genic maternal ability increased, however,
a leveling off occurred in the mating system analysis and a non-
significant depression in the sire testing analysis occurred when
PHCS changed from 50 to 80%.

The two higher levels of PHCS did not differ significantly from
one another in the amount of ADG improvement, however, both were
significatnly superior to 20% PHCS in phenotypic and genic value
averages. The 20% saving rate had 19% more phenotypic variance
and 34% more genic variance than 50 and 80% PHCS at calf crop 10 of

the mating system analysis.

116

The unadjusted WW was higher for the 20% PHCS level through
calf crop 7. However, by calf crop l0 enough improvement in direct
genic value for WW and maternal ability occurred in the 50 and
80% saving rate populations so that their unadjusted WW averages
were superior to the 20% level. The unadjusted YW averages were
significantly higher for 50 and 80% levels of PHCS than for 20% at
calf crop 7.

PHCS levels had a major effect on the generation interval. At
calf crop 10 in the mating system analysis, 20, 50 and 80% heifer
saving rates averaged 6.l, 3.7 and 2.8 calves per cow, respectively,
which gave calculated cow turnover rates of 7.6, 5.1 and 4.l years,
respectively, for 20, 50 and 80% PHCS.

Selection by YW or EBV caused the most improvement in YW. EBV
selection consistently gave greater improvement than YW selection
through calf crop 7. But by calf crop lo, YW selection gave heavier
calf YW, however, neither was significantly different from the other.

A possible, serious bias in female EBV was pointed out. The EBV
of a selected female is biased upward in herds that are rapidly
improving in performance. This occurs because the females' individual
deviation from herd mean was not readjusted, plus the fact that the
records for calves produced in the early part of a cows' productive
life are included in the EBV calculation as deviations from herd means
that are below the herd genetic mean at the time of dam selection.

YW selection caused the greatest amount of correlated improvement

in genic value for WW. However, it was not significantly greater than

117

EBV selection in either phenotype or genic value for WW. YW
selection also gave the greatest amount of improvement in genic
maternal ability, with EBV selection slightly below it.

ADG selection caused the greatest improvement in ADG to occur.
ADG selected populations for both analyses averaged .19 lbs. per
day higher than EBV and YW selected populations. EBV and YW
selected populations did not differ significantly in their improve-
ment rates for ADG.

Many interactions between the independent variables occurred,
however, five of the seven discussed in the thesis involved yearling
weight heritability (YWH). One of these was the interaction of
mating system and YWH for ADG in calf crops 7 and l0. For YWH
levels 20 and 40%, ADG was equal in RM and PAM, however, with 60%
YWH, ADG averaged .07 lbs. per day higher in PAM than RM.

Sire testing interacted with YWH in calf crop l0 for both ADG
and YW. A greater increase in both ADG and YW occurred as YWH
increased for the case of within purebred herd sire testing than for
sire testing in a commercial test herd. Sire testing also interacted
significantly with PHCS for YW and unadjusted YW in calf crop l0.
The interaction was the result of within purebred herd sire testing
responding to increased PHCS levels more than sire testing within a
commercial test herd.

YWH interacted significantly with selection criteria in WW, ADG
and unadjusted WW in both analyses and in all three calf crops

analyzed. Also it occurred for genic maternal ability in calf crOps

118

7 and lo of the mating system analysis. The cause of the interaction
was that ADG selected p0pu1ations did not respond as much to YWH
increases as EBV and YW selected populations for traits mentioned.
This interaction also occurred in the mating system analysis for ADG.
ADG selection gave more improvement in ADG than either EBV or YW
selection as YWH increased.

The interaction of YWH and PHCS was significant for YW and ADG
in all three calf crOps of the mating system data. Improvement
increased linearly across PHCS levels with 20 and 40% YWH, however,
with the 60% level of YWH, a leveling off occurred in ADG and YW
when PHCS changed from 50 to 80%. Significance of the interaction,
with the same action as described in the previous sentence, occurred
in genic value for WW in calf crOps 7 and lo, but only in calf crop
7 for WW phenotype.

The final independent variable YWH interacted with was bull
combination. This interaction influenced ADG in the mating system
analysis. At low YWH more improvement in ADG was made when more
progeny tested bulls were incorporated into the breeding system.
However, at the highest YWH level the use of fewer progeny tested
bulls enhanced the rate of improvement.

The last interaction discussed was between selection criteria
and PHCS. It was significant for ADG in the mating system data.

EBV and YW selected populations had an average improvement in ADG
of .08 lbs. per day when PHCS increased from 20 to 80%, while ADG
selected populations improved .20 lbs. per day as PHCS went from

20 to 80%.

APPENDIX A

Method gf_Calculating CWGl and CWGZ

 

CWGl and CWGZ are multiplicative factors used in BPAM and BTHS
to give a genetic correlation of 0.25 between WW and ADG. CWGl, as
used in this simulation, is equal to the covariance between WW and
ADG at the specified heritabilities divided by the genic standard
deviation for WW. CWGZ is then equal to the square root of the genic
variance in ADG minus CWGl squared. '

What follows next will be a step-by-step display of how CWGl and
CWGZ were calculated when heritability of yearling weight was equal

to 0.2. First formulas and then a numerical example.

 

 

FORMULAS:
1) °WW,ADG ‘ YWW,ADG '“ as" - ‘fioo
2) CWGl = auw,ADG / aww

3) CWGZ = ¢;ADG _ (CWG112

NUMBERICAL EXAMPLE:

 

1) 0WW,ADG ‘ 0-25 ”(1537043)z - (.13304)2

2) CW6] 0.51123 / 15.37043

 

3) cwez = “(.13304)2 - (.03326)2

119

BIBLIOGRAPHY

Bereskin, B., C. E. Shelby and L. N. Hazel. 1969. Monte Carlo
studies of selection and inbreeding in swine. I. Genetic
and phenotypic trends. J. Animal Sci. 29:678.

Bereskin, B., C. E. Shelby and L. N. Hazel. 1970. Monte Carlo
studies of selection and inbreeding in swine. II. Inbreeding
coefficients. J. Animal Sci. 30:681.

Bereskin, B. 1972. Monte Carlo studies of selection and inbreeding
in swine. IV. Selection response. J. Animal Sci. 34:726.

Blackwell, R. L., J. H. Knox, C. E. Shelby and R. T. Clark. 1962.
Genetic analysis of economic characteristics of young Hereford
cattle. J. Animal Sci. 21:101.

Blair, P. V., W. H. Kyle, V. L. Anderson and A. E. Bell. 1961.
Quantitative consequences of matin systems and selection
methods. Genetics 46:852. (Abstr.1

Brinks, J. S., R. T. Clark, N. M. Kieffer and J. R. Queensberry.
1962. Genetic and environmental factors affecting performance
traits of Hereford bulls. J. Animal Sci. 21:777.

Brinks, J. S., R. T. Clark, N. M. Kieffer and J. J. Urick. 1964.
Estimates of genetic, environmental and phenotypic parameters
in range Hereford females. J. Animal Sci. 23:711.

Brinks, J. S. 1965. Evaluation of response to selection and
inbreeding in a closed line of Hereford cattle. USDA Tech.
Bull. 1323.

Brown, C. J. 1958. Heritability of weight and certain body
dimensions of beef calves at weaning. Bull. 597, University
of Arkansas, Fayetteville.

Brown, C. J. and M. Gacula. 1964. Estimates of heritability of
beef cattle performance traits by regression of offspring on
sire. J. Animal Sci. 23:321.

Carter, R. C. and C. M. Kincaid. 1959a. Estimates of genetic and
phenotypic parameters in beef cattle. II. Heritability
estimates from parent-offspring and half-sib resemblances.

J. Animal Sci. 18:323.

120

121

Carter, R. C. and C. M. Kincaid. 1959b. Estimates of genetic and
phenotypic parameters in beef cattle. III. Genetic and
phenotypic correlations among economic characters. J. Animal
Sci. 18:331.

Chapman, A. B. 1951. Effectiveness of selection in laboratory
animals. J. Animal Sci. 10:3.

Chapman, H. 0., T. M. Clyburn and W. C. McCormick. 1969. Selection
of beef cattle for single traits. J. Animal Sci. 29:225.

Chapman, H. D., T. M. Clyburn and W. C. McCormick. 1972. Comparison
of criteria for selecting introduced beef sires. J. Animal
Sci. 35:321.

Cox, D. 1., J. E. Legates and C. C. Cockerham. 1958. Maternal
influence on body weight. J. Animal Sci. 18:519.

Cunningham, E. P. and C. R. Henderson. 1965. Estimation of genetic
and phenotypic parameters of weaning traits in beef cattle.
J. Animal Sci. 24:182.

Dalton, D. C. and T. L. Bywater. 1963. The effect of selection for
litter size and litter weight at weaning in mice maintained on
two diets. Animal Prod. 5:317.

Dalton, D. C. 1967. Selection for growth in mice on two diets.
Animal Prod. 9:425.

Deaton, 0. W. and L. D. McGilliard. 1965. Weighting information
from relatives to select for milk in Holsteins. J. Dairy Sci.
48:365.

Deese, R. E. and M. Koger. 1967. Maternal effects on preweaning
growth rate in cattle. J. Animal Sci. 26:250.

Dempster, E. R., S. M. Lerner and D. C. Lowry. 1952. Continuous
selection fOr egg production in poultry. Genetics 37:693.

Dickerson, G. E. 1949. Effectiveness of selection for economic
characters in swine. J. Animal Sci. 10:12.

Dickerson, G. E., C. T. Blunn, A. B. Chapman, R. M. Kottman, J. L.
Krider, E. J. Warwick, and J. A. Whattey, Jr. 1954. Evaluation
of selection in developing inbred lines of swine. Research
Bull. 551, University of Missouri, Columbia.

122

Enfield, F. D., R. E. Comstock and 0. Braskerud. 1966. Selection
for pupa weight in Tribolium Castaneum. 1. Parameters in base
populations. Genetics 54:523.

Falconer, D. S. 1960. Introduction to quantitative genetics. The
Ronald Press 00., New York, N.Y.

Fraser, A. S. 1957a. Simulation of genetic systems by automatic
digital computers. I. Introduction. Aust. J. Biol. Sci.
10:484.

Fraser, A. S. 1957b. Simulation of genetic systems by automatic
digital computers. II. Effects of linkage on rate of advance
under selection. Aust. J. Biol. Sci. 10:492.

Gill, J. L. 1965a. Effects of finite size on selection advance in
simulated genetic populations. Aust. J. Biol. Sci. 18:599.

Gill, J. L. 1965b. A Monte Carlo evaluation of predicted selection
response. Aust. J. Biol. Sci. 18:999.

Gill, J. L. 1965c. Selection and linkage in simulated genetic
populations. Aust. J. Biol. Sci. 18:1171.

Gill, J. L. 1966. Effects of selection and linkage on degree of
inbreeding. Aust. J. Biol. Sci. 19:307.

Hamann, H. K., S. Wearden and W. H. Smith. 1963. Estimation of
genetic and environmental factors affecting weaning weights
of creep-fed cattle. J. Animal Sci. 22:316.

Hetzer, H. 0. and J. H. Zeller. 1956. Selection for high and low
fatness in Duroc and Yorkshire swine. J. Animal Sci. 15:1215.

Hetzer, H. 0. Leanness made to order. Agricultural Research Service
Bull. (May, 1960):3-5.

Hetzer, H. 0. Seventh-generation leanness. Agricultural Research
Service Bull. (May, 1963):10-ll.

Hohenboken, W. D. and J. S. Brinks. 1971. Relationships between
direct and maternal effects on growth in Herefords: II.

Partitioning of covariance between relatives. J. Animal Sci.
32:26.

123

Hoornbeck, F. K. and R. Bogart. 1966. Amount of selection applied
and response of traits in four inbred lines of beef cattle.
Tech. Bull. 96. Oregon State University, Corvallis.

Kincaid, C. M. and R. C. Carter. 1958. Estimates of genetic and
phenotypic parameters in beef cattle. I. Heritability of
growth rate estimated from response to sire selection.~ J.
Animal Sci. 17:675.

Knapp, B. Jr., R. W. Phillips, W. H. Black and R. T. Clark. 1942.
Length of feeding period and number of animals required to

measure economy of gain in progeny tests of beef bulls. J.
Animal Sci. 1:285.

Knapp, B., Jr. and R. T. Clark. 1951. Genetic and environmental
correlations between weaning scores and subsequent gains in
the feed lot with Record of Performance steers. J. Animal
Sci. 10:365.

Knapp, B., Jr. and R. R. Woodward. 1951. Heritability of weights
of beef steers by months during the feeding period. J. Animal
Sci. 10:1026. (Abstr.)

Koch, R. M. and R. T. Clark. 1951. Genetic and environmental
relationships among economic characters in beef cattle. II.

Correlations between offspring and dam and offspring and sire.
J. Animal Sci. 10:786.

Koch, R. M. and R. T. Koch. 1955. Genetic and environmental
relationships among economic characters in beef cattle. III.
Evaluating maternal environment. J. Animal Sci. 14:979.

Koch, R. M. 1972. The role of maternal effects in animal breeding:
VI. Maternal effects in beef cattle. J. Animal Sci. 35:1316.

Kyle, W. H. and A. B. Chapman. 1953. Experimental check of the
effectiveness of selection for a quantitative character.
Genetics 38:421.

Li, C. C. 1955. P0pu1ation genetics. The University of Chicago
Press, Chicago, Ill.

Lush, J. L. 1935. Progeny test and individual performance as
indicators of an animal's breeding value. J. Dairy Sci.
18:1.

Lush, J. L., H. W. Norton, III, and F. Arnold. 1941. Effects
which selection of dams may have on sire indexes. J. Dairy
Sci. 24:695.

III I 11111 I

124

Lush, J. L. 1945. Animal Breeding Plans, 3rd. ed. Ames, Iowa:
The Collegiate Press, Inc.

Lush, J. L. 1947. Family merit and individual merit as bases for
selection. Part I. The American Naturalist 81:241.

Lush, J. L. 1948. The genetics of p0pu1ations (Mimeo.) Dept. of
Animal Science, Iowa State University, Ames.

McBride, G. and A. Robertson. 1963. Selection using assortative
mating in Drosophila melanogaster. Genet. Res. 4:356.

Minyard, J. A. and C. A. Dinkel. 1965. Heritability and repeat-
ability of weaning weight in beef cattle. J. Animal Sci.
24:1072.

Nelms, G. E. and P. 0. Stratton. 1967. Selection practiced and
phenotypic change in a closed line of beef cattle. J. Animal
Sci. 26:274.

Nielsen, M. K. 1974. An examination of analysis procedures for
national beef sire evaluation programs. Ph.D. Thesis, Iowa
State University, Ames.

Pahnish, 0. F., E. B. Stanley, R. Bogart and C. B. Roubicek. 1961.
Influence of sex and sire on weaning weights of southwestern
range calves. J. Animal Sci. 20:454.

Pahnish, 0. F., R. L. Roberson, R. L. Taylor, J. S. Brinks, R. T.
Clark and C. B. Roubicek. 1964. Genetic analysis of economic
traits measured in range-raised Herefords at preweaning and
weaning ages. J. Animal Sci. 23:562.

Parker, R. J., L. D. McGilliard and J. L. Gill. 1969. Genetic
correlation and response to selection in simulated populations.
I. Additive model. Theoret. Appl. Genet. 39:365.

Rahnefeld, G. W., W. J. Boylan, R. E. Comstock and M. Singh. 1963.
Mass selection for post-weaning growth in mice. Genetics
48:1567.

Rollins, W. C. and K. A. Wagnon. 1956. A genetic analysis of wean-
ing weights in a range beef herd operated under Optimum and
sub-optimum nutritional regimes. J. Animal Sci. 15:125.

125

Ronningen, K. 1969. Studies on selection in animal breeding. I.
The efficiency of two-stage selection compared with single-
stage selection with respect to progeny testing in animal
breeding. Acta Agr. Scand. 19:149.

 

Ronningen, K. 1970. Studies on selection in animal breeding.
II. The efficiency of selection based on the combination
of performance testing and progeny testing compared with
selection based only on performance testing when testing
capacity is limited. Acta Agr. Scand. 20:71.

 

Shelby, C. E., R. T. Clark, J. R. Queensberry and R. R. Woodward.
1960. Heritability of some economic characteristics in
record of performance bulls. J. Animal Sci. 19:450.

Shelby, C. E., W. R. Harvey, R. T. Clark, J. R. Queensberry and
R. R. Woodward. 1963. Estimates of phenotypic and genetic
parameters in ten years of Miles City R.0.P. steer data.

J. Animal Sci. 22:346.

Stanley, E. B. and R. McCall. 1945. A study of performance in
Hereford cattle. I. Progeny testing of Hereford sires.
Tech. Bull. 109. University of Arizona, Phoenix.

Sutherland, T. M., P. E. Biondini and L. H. Haverland. 1968.
Selection under assortative mating in mice. Genet. Res.
11:171.

Swiger, L. A. 1961. Genetic and environmental influences on gain
of beef cattle during various periods of life. J. Animal
Sci. 20:183.

Swiger, L. A. and L. N. Hazel. 1961. Optimum length of feeding
period in selecting for gain of beef cattle. J. Animal
Sci. 20:189.

Swiger, L. A., R. M. Koch, K. E. Gregory and V. H. Arthaud. 1962.
Selecting beef cattle for economical gain. J. Animal Sci.
21:588.

Taylor, J. C., R. C. Carter, C. M. Kincaid, 8. M. Priode and J. A.
Gaines. 1960. Estimates of genetic and phenotypic parameters
in beef cattle. IV. Repeatability of cow perfbrmance. J.
Animal Sci. 19:700.

126

Vesely, J. A. and 0. W. Robison. 1971. Genetic and maternal effects
on preweaning growth and type score in beef calves. J. Animal
Sci. 32:825.

Wagnon, K. A. and W. C. Rollins. 1959. Heritability estimates of
post-weaning growth to long yearling age of range beef heifers
raised on grass. J. Animal Sci. 18:918.

Willham, R. L. 1963. The covariance between relatives for characters
composed of components contributed by related individuals.
Biometrics 19:18.

Willham, R. L. 1972. The role of maternal effects in animal :
breeding: III. Biometrical aspects of maternal effects in r
animals. J. Animal Sci. 35:1288.

 

Willham, R. L. 1973. Breeding Values. Simmental Shield, May, 1973,
pp 0 52-59 0

Wilson, S. P., W. H. Kyle and A. E. Bell. 1965. The effects of
mating systems and selection on pupa weight in Tribolium.
Genet. Res. 6:341.

Wilson, S. P., W. H. Kyle and A. E. Bell. 1966. The influence of
mating systems on parent-offspring regression. J. Hered.
57:124.

Wright, S. 1921a. Systems of mating. I. The biometric relations
between parent and offspring. Genetic 6:111.

Wright, S. 1921b. Systems of mating. III. Assortative mating
based on somatic resemblance. Genetics 6:144.

Wright, S. 1921c. Systems of mating. V. General considerations.
Genetics 6:167.

Young, C. W. and J. E. Legates. 1965. Genetic, phenotypic, and
maternal interrelationships on growth in mice. Genetics
52:563.

MICHIGAN STATE UNIV. LIBRARIES
llHlHWWIIWIW11111111N1111111111111111
31293006587178