 

A2
5-9.“. .3“...
1.1,...

 

 

$25.3 ﬁt ‘ ‘ . Lam“ Jam.“ £3. _....r. ‘ V . . anmwu“...a ., Maud...

 

 

THESIS

A t“ '
.1 l

(l-

‘x
\J

Date

0-7 639

”mm liliﬂliilljllliliﬂljlln

3 1293 01

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

thesis entitled

ANALYSIS OF GENETIC PARAMETERS FOR
GROWTH AND CARCASS TRAITS 0F

CANADIAN

CHAROLAI S CATTLE .

presented by
DWIGHT AARON SEXTON

has been accepted towards fulﬁllment

of the requirements for

M. S. degree in WIENCE

 

MAY 10. 1996

 

A759”; 43%,

 

Major professor

MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

 

' I t *_ _._ ‘— ' ~—

-————— - r‘ — v f v w“--

a.» A '— v -

- 'f-V' v—-

 

PLACE IN RETURN BOX to remove We checkout from your record.
TO AVOID FINES return on or before date due.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

D
L
E

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ALﬂ
I T—Ir—IF

 

 

 

 

MSU lo An Afﬁrmative ActionlEqual Opportunity Institution

smarts-9.1

 

ANALYSIS OF GENETIC PARAMETERS FOR GROWTH AND
CARCASS TRAITS OF CANADIAN CHAROLAIS CATTLE

By

DWIGHT A. SEXTON

A THESIS

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

MASTER OF SCIENCE

Department of Animal Science

1996

ABSTRACT

Analysis of Genetic Parameters for Growth and
Carcass Traits of Canadian Charolais Cattle.
By

Dwight A. Sexton

Growth and carcass records were received from the Canadian Charolais Association’s
Conception to Consumer Program which encompassed 5401 progeny of 172 sires, born
from 1975 through 1993. Sires were randomly bred to mature commercial cows, and the
offspring were placed in a commercial feedlot and a 112-day performance test. Progeny
of sires were slaughtered when they visually reached the A1 or A2 Canadian carcass
grade. Carcass weight, longissimus muscle area, 12th rib backfat, marbling score, carcass
grade, and cutability were analyzed using slaughter age as a covariate. In a separate
analysis, longissimus muscle area, 12th rib backfat, and marbling score were analyzed
using carcass weight as a covariate. A ﬁve trait sire model that included sire relationships
was employed using an average information REML (DMUAI) algorithm to estimate
(co)variance components. Heritability estimates for the following traits included: birth
weight, 0.22; adjusted 200 day weight, 0.11; adjusted 365 day weight, 0.19; post-weaning
average daily gain, 0.21; end of test weight, 0.18; slaughter age adjusted carcass weight,
0.13; slaughter age adjusted marbling score, 0.27; carcass weight adjusted marbling score,
0.28; slaughter age adjusted longissimus muscle area, 0.29; carcass weight adjusted
longissimus muscle area, 0.38; slaughter age adjusted 12th rib backfat thickness, 0.37;

carcass weight adjusted 12th rib backfat thickness, 0.36; slaughter age adjusted carcass

carcass weight adjusted 12th rib backfat thickness, 0.36; slaughter age adjusted carcass
grade, 0.23; slaughter age adjusted cutability, 0.32. Genetic (rs) and phenotypic (rp)
correlations between the growth traits analyzed in most instances were moderate to high.
With the exception of carcass weight, the (re) and (rp) of grth by carcass traits were low
to moderate. When adjusted to an age constant basis, the estimated genetic correlation
(r5) between carcass weight and longissimus muscle area was 0.18, while the estimate of
(r,) between carcass weight and 12th rib backfat thickness was 0.17. The (r3) between
marbling score (increased marbling had a lower numerical score) with longissimus muscle
area and also to 12th rib backfat thickness was 0.16 and -0.32, respectively. The effect of
selection on the females had little, if any, impact on the estimation of genetic parameters.
These data indicate that successful selection for growth and carcass traits can occur, but

antagonistic results may transpire.

DEDICATION

In loving memory of my father, Larry H. Sexton, whose greatest joy in life was the

success of his four children. I’ll keep trying to make you proud.

iv

ACKNOWLEDGMENTS

The author is grateful to Dr. Dennis Banks for his guidance, insight and support
throughout my program. Additionally, Dr. Al Booren, Dr. Dave Hawkins, and Dr. Harlan
Ritchie have been great sources of inspiration to me. The support of everyone on my
graduate committee has not only made me a better scientist, but also a better steward of
the livestock industry.

Additionally, this project was only possible due to contribution of the Canadian
Charolais Association for providing the data, and also to Dr. Ivan Mao and Dr. Just

Jensen for making their computing resources available.

TABLE OF CONTENTS

List of Tables .............................................................................................................. viii
1 Introduction ............................................................................................................. 1
2 Review of Literature ................................................................................................ 3
2.1 Introduction .................................................................................................. 3

2.2 Genetic Aspects Of Growth And Carcass Traits ............................................ 4

2.2.1 Birth Weight .................................................................................... 4

2.2.2 Weaning Weight .............................................................................. 7

2.2.3 Yearling Weight ............................................................................. 15

2.2.4 Post-weaning Average Daily Gain .................................................. 16

2.2.5 End of Test Weight ........................................................................ 17

2.2.6 Carcass Weight .............................................................................. 18

2.2.7 Longissimus Muscle Area .............................................................. 20

2.2.8 Marbling ........................................................................................ 21

2.2.9 Twelfth Rib Fat Thickness .............................................................. 22

2.2.10 Carcass Grade ................................................................................ 24

2.2.1 1 Cutability ....................................................................................... 24

2.3 Parameter Estimation .................................................................................. 26

2.3.1 Model Speciﬁcation ....................................................................... 26

2.3.2 Single and Multiple Trait Mixed Model Methodology .................... 28

2.3.3 (Co)variance Estimation ................................................................. 29

2.3.4 Genetic Parameter Estimation ........................................................ 31

2.3.4.1 Heritability Estimators ....................................................... 31

2.3.4.2 Correlation Estimators ...................................................... 32

3 Materials and Methods ............................................................................................. 34
3.1 Description of the Data and Data Edits ....................................................... 34

3.2 Estimation of Genetic Parameters ................................................................ 39

3.2.1 Model 1: Single Trait Mixed Model .............................................. 39

3.2.2 Model 2: Single Trait Mixed Model .............................................. 41

3.2.3 Model 3: Five Trait Mixed Model ................................................. 47

3.2.4 Model 4: Four Trait Mixed Model ................................................. 54

4 Results and Discussion ............................................................................................. 62
4.1 Model 1 Genetic Parameters ....................................................................... 62

4.2 Model 2 Genetic Parameters ....................................................................... 62

4.3 Model 3 Genetic Parameters ....................................................................... 65

4.3.1 Heritability Estimates ..................................................................... 74

4.3.2 Genetic Correlations ...................................................................... 82

4.3.3 Environmental Correlations ............................................................ 86

vi

4.3.4 Phenotypic Correlations ................................................................. 87

4.4 Model 4 Genetic Parameters ........................................................................ 91

4.4.1 Heritability Estimates ..................................................................... 91

4.4.2 Correlation Estimates ..................................................................... 97

5 Conclusions ............................................................................................................. 99
6 Literature Cited ...................................................................................................... 101

vii

Table 1.

Table 2.

Table 3.

Table 4.

Table 5.

Table 6.

Table 7.

Table 8.

Table 9.

Table 10.

Table 11.

Table 12.

Table 13.

Table 14.

Table 15.

Table 16.

Table 17.

LIST OF TABLES

Mean and range of literature heritability estimates ..................................... 8
Average and range of genetic correlations among grth traits. ............... 9
Average and range of genetic correlations among carcass traits. .............. 10

Average and range of genetic correlations among grth and carcass

traits. ...................................................................................................... 11
Average and range of phenotypic correlations among growth traits. ....... 12
Average and range of phenotypic correlations among carcass traits. ........ 13

Average and range of phenotypic correlations among grth and

carcass traits. .......................................................................................... 14
Data editing criteria and number of records deleted. ............................... 37
Decrease in volume of records by sex. .................................................... 38
Number of records after edits, means, and standard deviations ................ 42

Model 1 estimates of sire variance, residual variance, and heritability ...... 63

Model 2 estimates of heritability and sire, residual, and phenotypic

variance .................................................................................................. 64
Model 3 genetic variances of ﬁve trait combination subsets ..................... 66
Model 3 environmental variances of ﬁve trait combination subsets .......... 67
Model 3 phenotypic variances of ﬁve trait combination subsets. .............. 68
Model 3 heritability estimates of ﬁve trait combination subsets ................ 69

Model 3 mean, standard deviation, and range of genetic variance
estimates ................................................................................................. 70

viii

Table 18.

Table 19.

Table 20.

Table 21.

Table 22.

Table 23.

Table 24.

Table 25.

Table 26.

Table 27.

Table 28.

Model 3 mean, standard deviation, and range of environmental variance
estimates ................................................................................................. 71

Model 3 mean, standard deviation, and range of phenotypic variance

estimates ................................................................................................. 72
Model 3 mean, standard deviation, and range of heritability estimates ..... 73
Model 3 average genetic and environmental correlations for growth

traits ....................................................................................................... 79
Model 3 average phenotypic correlations for growth traits ...................... 80

Model 3 average genetic and environmental correlations for carcass
traits ...................................................................................................... 81

Model 3 average phenotypic correlations for carcass traits ...................... 92

Model 3 average genetic and environmental correlations between
growth and carcass traits ........................................................................ 93

Model 3 average phenotypic correlations between grth and carcass
traits ....................................................................................................... 94

Model 4 heritability, genetic variance, environmental variance, and

phenotypic variance estimates ................................................................. 95
Model 4 genetic, environmental, and phenotypic correlations .................. 96

ix

1. INTRODUCTION

When the continental breeds of cattle were introduced into North America in the
1950’s through the 1970’s, a change in traits of economic importance under selection also
occurred. Fat was no longer considered desirable, and leaner, faster growing cattle were
desired. Additionally, through the inclusion of these later maturing breeds, commercial
cattlemen were able to reap the rewards of selection, migration, and heterosis at a much
faster pace. Furthermore, during this same time period, great strides were made in the
ﬁeld of animal breeding; and animal scientists were much better equipped to identify
genetically superior animals through the use of mixed model methodology (Henderson,
1953)

In an effort to increase their market share in the beef industry, members of the
Canadian Charolais Association designed an unbiased progeny test program that enabled
bulls to be proven for growth and carcass traits. There was selection on the heifers used
in the study, which may result in selection bias if it was not a random culling of the heifers.
As a result of this program, data were available for the estimation of (co)variance
components through the use of multiple trait mixed model methodology. Heritability
values, in addition to genetic, environmental and phenotypic correlations, could be
estimated. Heritability estimates enable producers to explain variation, or the lack of
variation from parent to progeny, when designing selection programs. Genetic
correlations give an indication of how traits not directly selected upon will change when
producers use estimates of genetic merit when making their mating decisions.

Environmental correlations are important to note in selection programs because they can

either have antagonistic or desirable effects that will eﬁ‘ect the phenotypic correlation.
Phenotypic correlations are indicators of how traits will react together if they are not
selected for with breeding values. With these statistics, members of the Canadian
Charolais Association can measure the degree of potential progress possible through
selection for the traits evaluated in the Conception to Consumer Program. In addition to
what this information provides to individual breeders, these details are important to the
Canadian Charolais Association as they continually update their sire summaries because
they need prior values to use for their assessment of expected progeny differences.

The traits evaluated in this study include: birth weight, age adjusted 200 day
weight, age adjusted 365 day weight, post-weaning average daily gain on test, and end of
test weight. Additional carcass traits involved in this study include: hot carcass weight,
longissimus muscle area, 12th rib backfat thickness, marbling score, carcass grade, and
also cutability percentage.

The overall objective was to study progeny from 172 Charolais sires to provide
heritability estimates on various traits. The speciﬁc objectives of this study were to:

1. Estimate the genetic and phenotypic parameters among growth and

carcass traits.

2. Explore the potential selection bias from the selection on females in

these data.

2. LITERATURE REVIEW

2.1 Introduction

Genetic estimates for growth and carcass traits in beef cattle provide an indication
of the progress that can be made by selecting for certain traits and the resulting change in
correlated traits. Falconer (1960) describes methods to show advances in the genetic

makeup of livestock. The formula necessary to estimate genetic progress is as follows:

Agzi—hjﬂa.
GI

where: Ag is genetic progress,
1' is the notation for selection intensity, which is the “standardized”

selection diﬁ‘erential of normally distributed traits,

\[h—f is the square root of heritability (the accuracy of the breeding value
based on individual phenotypic records),
a .4 is the additive genetic standard deviation for the trait, and
GI. is the generation interval for the population.
Additionally, Falconer (1960) describes the measurement of the correlated response to
selection, which is the response in a second trait that occurs after selection has occurred

upon the initial trait. The formula for the correlated response to selection is:

C.R.,,= ile‘ll—rggva,

where: C.R.y is the correlated response of trait Y when trait X is selected for,

iis the notation for selection intensity, which is the “standardized”

selection differential for the normal distribution,
‘/ h X2 is the square root of heritability (the accuracy of the breeding value

based on individual phenotypic records), for the selected
trait X,

W is the square root of heritability (the reliability of the phenotypic
value as a guide to the breeding value) for the correlated

trait Y,

’23., is the genetic correlation between traits X and Y,

O'y is the phenotypic standard deviation of trait Y, and

GI. is the generation interval for the population.

Because of their importance to genetic evaluations, estimates of the genetic and
phenotypic parameters are essential to the future of any and all populations undergoing

selection.

2.2 Genetic Aspects Of Growth And Carcass Traits

2.2.1 Birth Weight
The birth weight of beef calves is of critical interest to beef cattle producers
(B.I.F., 1990) as lighter weight calves tend to have lower mortality rates, are born easier,

and result in less rebreeding diﬁiculties for the dam. Koots et al. (19943), in a paper

which summarized published genetic parameters of 287 papers from North America and
Europe analyzing 70 traits, reported 172 birth weight heritability values had a mean
heritability of 0.35 with a 0.16 standard deviation. Koots et al. reported that the
heritability values were affected by the mean and phenotypic standard deviation of the
population, in addition to the effects of breed, sex, method of parameter estimation,
feeding management, and data origin, although the magnitude and direction of the eﬂ‘ects
varied. Moreover, Koots et al. noted that traits with low heritability values tended to have
the average heritability value overestimated. Johnston et al. (1992) used a two trait sire
model without sire relationship information on 1444 Charolais sired progeny from the
Canadian Charolais Association’s Conception to Consumer Program to report a birth
weight heritability estimate of 0.25. Koch et al. (1982) analyzed data from 2,453 steers at
the Germ Plasm Evaluation project at the Meat Animal Research Center to estimate a
birth weight heritability of 0.43. de Rose (1992) estimated 0.45 to be the birth weight
heritability for Charolais in the Canadian Beef Sire Evaluation Program using a multiple
trait animal model. Woodward et al. (1992) reported an estimate of birth weight
heritability on 13,670 Sirnmental progeny to be 0.28 (Table 1). Heritability values of this
magnitude suggest that genetic progress can be made when selecting for lower birth
weights.

The genetic correlations reported in the literature of birth weight to other growth
traits are represented in Table 2. The mean genetic correlations of birth weight and other
growth traits (weaning weight, yearling weight, post-weaning average daily gain, and end

of test weight) reported ﬁ'om a review of the literature cited were 0.45, 0.48, 0.43, and

0.41, respectively. Table 4 contains the average and range of literature genetic
correlations for grth and carcass traits. The average genetic correlation between birth
weight and carcass weight was 0.44, while the average correlation between birth weight
and longissimus muscle area was 0.40. These growth and carcass trait correlations
suggest that selection for lower birth weights would result in lighter weight cattle with
smaller longissimus muscle areas. Additional genetic correlations of birth weight to 12th
rib fat thickness, marbling score, and cutability included respective mean values of -0.27,
0.12, and 0.10. Therefore, when selecting for lighter birth weight calves, the correlated
genetic response would yield carcasses with more fat, and lower cutability cattle with
more marbling.

The mean phenotypic correlations for each of the trait combinations with birth
weight possessed the same sign but had lower magnitudes than the respective genetic
correlations (Table 2; Table 4; Table 5; Table 7). Koots et al. (1994b) analyzed 66 and 42
citings in the literature and determined the mean phenotypic correlations of birth weight to
weaning weight and also birth weight to post-weaning average daily gain of 0.36 and 0.20,
respectively. Koots et al. (1994b) conducted a weighted least squares analysis of
literature estimates of each correlation and showed several factors signiﬁcantly (P<0.10)
affecting the estimates, including breed, country, sex, and decade in which data were
collected. Other factors such as data origin (ﬁeld data or experimental data), feeding
regime (range or feedlot) and estimation method such as sire versus animal model and
single versus multiple trait analysis generally did not signiﬁcantly affect genetic and

phenotypic correlations.

2.2.2 Weaning Weight

The weaning weight of a calf is the best measure of pre-weaning growth. Arnold
et al. (1991) reported a weaning weight heritability of 0.09 from a study of 2411 Hereford
steers ﬁom the American Hereford Association’s sire evaluation program which used a
two trait sire model in the analysis. de Rose (1992) used data from the Canadian Beef Sire
Evaluation Program on Charolais and Charolais-sired cattle to estimate a heritability value
of 0.25 for weaning weight with a multiple trait, animal model. Woodward et al. (1992),
Nunez-Dominguez et al. (1993 ), and Veseth et al. (1993) reported heritability values of
0.18, 0.37, and 0.17, respectively, for weaning weight (Table 1). Johnston et al. (1992)
also reported a weaning weight heritability value of 0.09 as did Arnold et al. (1991), but
also cited Robertson (1977), noting that if selection of the parents is based on the trait on
which heritability is being measured then the estimates may be biased due to reduced
additive genetic variance of the parents. Koots et al. (1994a) summarized 239 weaning
weight heritability estimates and found a mean of 0.27 with a standard deviation of 0.17.
These results indicate that weaning weight is low to moderately heritable and moderately
low genetic response to selection may be expected for this trait.

Weaning weight has been reported to have a medium to high genetic and
phenotypic correlation with other grth traits. Koots et al. (1994b) determined a mean
genetic correlation of 0.78 ﬁ'om 66 references and 0.72 for a mean phenotypic correlation
from 77 sources for weaning weight correlated to yearling weight. Marshall (1994) also
conducted a review of the recent scientiﬁc literature on breed characterization and genetic

parameters for beef cattle carcass traits and after averaging correlations of weaning weight

22.2 .222 a 22.22.228.22 .2
“482 .24 2... 2282 .2

482 22422232 .2

2.2 .22. s 22822? .2

82 .22. a .2208> .22

32 r? 3 nonwcgonéogz .

mag .omsonu—Bm .

82 .22. a 285 224> .

o
a
N82 :3 8 $832503 .w
h
o

mag .83! cu .

N32 :3 8 22322282. .m
282 .222 a 2.2252 .4
SE :3 2» 2260M .m
2222 322222822 .N
232 .222 8.322220 .2 23
222226228 Emma? mgoao 2.
2522.223 owe Simon—m a
.22—225 mam—23> 3222.822 ow< .2
2.222203 22:83 828.22; 022... .

 

 

2 .2 .22 .N .2 “no $.22 2 2 .o . 222222235

2 .42 .2 24.22 2.22 8 4.4.22 _. an.5222: 24.2 .2222 2222

42 .2 .2 .22 .4 .22 .N .2 24.0 «no 8 8.22 .. 305202222. 3.2 22222 2222
2 .42 .2 224.0 0422 9 «no _. 8:. M282242 322222.293

42 .2 .2 .22 .5 .4 .m .N .2 4.22 8.22 9 N2 . 82 023:2 msémmaaﬁ
2 .2 .22 .2 24.22 2.22 9 223 _. 28m 222224322

2 .2 .2 .22 .o .22 .5 .4 .m .N .2 28 54.22 9 mad . 28m 2222.52
2 .2 .2 .22 .4 .m .2 224.22 8.22 3 43 2%? cacao

2 .22 .N 44.o «no 8 2.22 22222203 28.4.20 2m

2 .22 .m .4 .m .N one P3 9 2 .o 22240 2225 02203.. magseom

2 .2: .e .m .4 2.3 N42. 3 2 .o .2 222203 m=2228>

2 .22 .222 .22 .e .m .4 8.22 two 8 8.22 22222203 2222283

42 .22 .2 .22 .e .m .m 48 54.22 3 2 .22 22222203 2222222
89:25 :32 omen”— «22,—.

 

Sonata 52.22.32.22— 2528= be owe-2.. ES 2282 .2 03am.

 

.2422 .222 2» 382 .n

82 .24 s 2228.5 .4

82 .22. s 2.22.38; .m

2.22 .22. c 2.282 .N
22 .24 s 2282 .2 2.183822. 22
8.22 422.2 9 222 ﬁne? 28.2.8 2.22m .. .225 2228 omega. 3:83.22
8.22 1 222203 28.20 2.22m .. 2.2203 22228»
224.22 222.22 a 2 .22 2228 222.5 omega. 222222283282 .. 22222203 22222.3
422 2 .2 a 23 222203 282.20 2222 .. 222%? 2.22283
224.22 2.2 2 2.22. .2225 228 $234. 222283.282 .. 22222203 222228?
N42. 2222 3 8.22 22303 ”22222822. .. 22222203 «2222233
24.22 2.22 a 2222 is? auto 2am _. 22222203 2222222
m4... 222 a 2.22 .228 2228 omega. 2222.83.38 . 2.2203 222.222
224.22 i 222203 22222822. . awe; 2222222
n4... 422 a 2.22 25225 wees», _. 2222...? .2522
2.82 2.52 2.24.2;

 

 

6.2.2»th

 

 

.31.: again- guﬁheo owe—.0» no mus... eon ”22.293 .n 03:.

10

£43— :3 8 38M
33 43.3.32

82 :3 8 228:3
33 :3 «a 588/
«aa— :3 8 934683

xo'rx'oo'oio'

4442 .24 8 285 44>
2442 .24 8 42425.

4442 .24 24 22842

2442 .24 24 4.22445

2442 .24 8 484424244242

v-ioim'viori

21482228 .26 2422

4.24.4323 2232228 owe 282322424 4 2228.2 24 23.22 44828 =< 4

 

 

 

42 4 44.4. 44.4. 42 44.4. 422224425 4 4822242224 24.2 22242 22242

42 .4 44.4 44.4 42 44.4 4222222425 4 82 428422 442424424442

42 .4 .4 .4 .4 .2 44.4. 44.4. 42 44.4. 485242224 24.2 .2242 22242 4 8:. 42842 442424424442

2.4 .4 .4 44.4. 42.4. 42 44.2- 422224245 4 284 422224242

42 .4 .4 .4 .4 .2 44.4 44.4 42 2 .4. 4822242222. 2442 .2242 22242 4 284 44224242

42 .4 .4 .2. .4 .4 .4 .2 44.4. 24.4 42 44.4. 824. 428442 44228424442 4 284 422228422

42 .4 .4 22.4. 42.4 s 44.4. 42222224226 4 2224243 448240

42 .4 .4 .4 .4 .4 44.4 44.4 42 44.4 484442222. 2442 4242 22242 4 2.24243 4.48240

42 .4 .4 .4 .4 .4 .4 44.4 44.4 42 44.4 8:. 424442 442424424442 4 2224243 448240

42 .4 .4 .4 .4 .4 .4 44.4 44.4 42 44.4. 284 442244422 4 2224243 448240

824.44. .4844 445E 434.54.
cease

 

 

.. 53.5 443.28 use—54 enemas—0.2.23 cream” .3 mung .252 ""395. .n 03:.

4442 .24 24 44432283 .4
2442 .24 24 42442 .4
4442 242422842 .2 ”a

932 :3 40 $8M .o
33 ..32332 .n
82 :3 8 586> .v

11

24.43224 2232228 owe 28232224 4 2228.4 6.24 42322 44828 =< 4

 

 

4 44.4. 1 422224425 4 2.242444 282.20 4442

4 44.4 I 484242242. 2442 4242 2442 4 2.2424? 28.440 4442

4 .4 44.4 24.4 s 44.4 82 428442 442424424442 4 2.2483 28.2.20 22

4 .4 2.4.4 2.4.4 42 44.4 4284 442242442 4 2224243 244440 4442
4.4 24.2 44.2 42 44.4 224243 448240 4 224243 28.2.40 44m

4 44.4 r. 422224425 4 42444 422442 4442442 442483.282

4 .4 .2 2 .4 44.4 42 44.4 4822242242. 2442 4242 22242 4 42424 422422 444242 442483.282
4 .4 .4 .2 24.4 44.4 42 42 .4. 82 42842 422422442443 4 222424 422442 444242 442483-282
4 .4 .4 .2 44.4 44.4 2 44.4 284 442242442 4 22240 422442 4442442 442483.248
4 .4 .4 .2 44.4 22.2 3 44.4 224243 448240 4 42444 422442 44424.2 442483.282
4 .4 44.4 44.4 42 42 .4. 422224425 4 2224243 42222284.

4 .4 .4 22.4 44.4 42 42.4. 484242224 24.2 4222 22242 4 2224243 4.222284.
4 .4 .4 24.4 44.4 42 44.4. 82 428442 444224424442 4 2224243 44222844
4 .4 .4 44.4. 44.4 a 44.4. 284 442242442 4 2224243 44222824
4 .4 .4 44.4 44.4 42 44.4. 2.24243 44.8440 4 224243 44224824
4 .4 .4 44.4 44.4 42 44.4. 422224425 4 2224243 4422283
4 .4 .4 44.4 44.4 42 44.4. 4.8424222. 2442 4222 22242 4 224243 442483
4 .4 .4 .4 44.4 44.4 42 44.4 82 4282242 42222224424442 4 2224243 442483
4 .4 .4 .4 .4 44.4 24.4 42 42.4. 284 422224.422 4 2.24243 442483
4 .4 .4 .4 44.4 22.2 42 2.4 224243 4485 4 2.24243 4222483
4 .4 42 .4 42 .4 a 44.4 422224425 4 2224243 2242242

4 .2 44.4. 44.4. 42 44.4. 4822242224 24.2 4222 2242 4 224243 52242

.4 .2 44.4 2.4.4 42 24.4 82 42842 442222442443 4 2224243 52242

4. .4 .2 42.4 24.4 a 42.4. 284 442242442 4 224243 222242
4 .4 .2 44.4 44.4 8 22.4 2224243 8828 4 2224243 224242
.3»: ems 33.5:

 

g

 

 

.324... 448.28 4.... SS. 4444442248 44 44.. 24.... 4. .4 42.244

12

44442 .24 44 4.4432 .4

4442 .24 24 224845 .4

4442 .24 8 2224442483 .4
4442 4244422842 .2 ”ﬁg

 

 

4 .4 44.4 44.4 42 44.4 444243 .820 2.442 4 428 422442 4448.2 444483.444

4 .4 44.4 44.4 42 44.4 2224243 28.20 2444.2 4 4224243 442483

4 .4 44.4 44.4 42 42 .4 4240 422442 44482 442484444442 4 2.2443 442483

4 44.4 ... 4224243 44424844 4 2224243 45483

4 .4 24.4 44.4 4. 44.4 222483 284.20 2.442 4 2224443 22542

4 .4 .2 44.4 44.4 42 42 .4 44444 422442 44482 442.283.4442 4 4224243 22542

4 .4 .4 24.4 44.4 42 44.4 2224443 4422283 4 2224243 2224242

844.444. .4844 $5.: 4244.54.
4.444.444.4444

 

 

 

.4395 532M use—5. Beau—9:8 ﬂan 3 mug... ...—a HEQ>< .m 034,—.

13

4442 .24 24 222432283
4442 .24 24 22445 44>
4442 .24 24 22432

22.42 ..24 2424222440

2442 .24 24 42422424244242

5&3 4.33896“ .a

33 4:53.32 .w
83 :3 20228:? .h
32 :38580> .0

Acid-id

2l§8 .24 24242

424.435 2222424228 own 242232324 42 22292.2 24 423.22 44828 :< 4

 

 

44.4. ... 4222222425 4 485245 2442 .2222 22242
44.4 ... 42222224240 4 82 424442 44222244244442
42.4. 44.4. 42 44.4. 484242222. 2442 4222 22242 4 82 424422242 444224424442
44.4. 42.4. 42 44.4. 4222244226 4 4284 4222242422
44.4 44.4 42 42 .4 485242224 2442 .2222 22242 4 4284 44224242
44.4 42 .4 42 8.4. 82 42444242 44222244244442 4 4284 422224242
44.4. 44.4 42 44.4. 422224425 4 2224243 448240
44.4 44.4 42 44.4 48422422: 24.2 .2222 .2242 4 2224243 448240
44.4 44.4 42 44.4 82 424442 44222244244442 4 2224243 448240
42.4 44.4 42 44.4 284 44224242 4 2224243 448240
2.8422 4.4.322 4244.244;
42.444.44.242

 

 

 

.4443... 22223.23 WEE—u 42283—223 922.582.?! 30.5. .4 323,—.

14

33 :3 20 5040> .m
4.32.2— :3 20 88M .24. N82 :3 20 P243683 .N
32 4232—432 .44 S2 :3 2o 233— .2 2432228 mo 243

4222.43.24 222424228 ow4 2422322424 4 2228.2 84 422422 444240 22444 4

 

 

 

4 44.4. I 422224425 4 2224243 2820 22422

4 42 .4 ... 48222245 2442 4222 22242 4 2.222243 244.20 22442

4 .4 44.4 44.4 42 44.4 82 424442 44222244242242 4 2224243 282.20 222242

4 .4 42.4 44.4 42 42.4 4284 422224242 4 2224243 282.20 22224

4.4 44.4 44.4 42 44.4 22242425 4.48240 4 2224243 28420 222242

4 44.4 ... 422224426 4 222442 222442 444242 442483.442

4 .2 42 .4 42 .4 42 2.2 .4 4842245 2442 4242 22242 4 42422 422442 4224242 42222284224442

4 .4 .2 44.4 44.4 42 44.4 82 42442222 42222224424442 4 22240 422442 444242 22222484224442

4 .4 .2 42 .4 44.4 42 44.4 4284 422224.242 4 22240 422442 444242 42222284424442

4 .4 .2 44.4 42.4 42 44.4 222424? 4.4.8240 4 22240 422442 444242 442483.442

4 42 .4. ... 4222244226 4 2224243 422222824

4 44.4 ... 4442222425. 2442 4222 22242 4 22222243 44222824

4 44.4 ... 82 4244.222 4422224424442 4 2224243 44222824

4 42.4 ... 4284 42222222422 4 2224243 224222822.

4 44.4 ... 2224243 448244 4 22222243 422222824

4 .4 .4 44.4 44.442 42.4. 422222425 4 2224243 42222283

4 .4 42 .4 24.4 42 42 .4 485242224 2442 22222 22242 4 2224243 4222483

4 .4 .4 44.4 44.4 42 44.4 82 42442232 42242244244442 4 2224243 222222283

4 .4 . . 44.4 42.4 42 44.4. 4284 42222222442 4 2224243 222222283

4 .4 .4 44.4 44.4 42 44.4 222424? 444428 4 2224243 4222483

4 .4 44.4 44.4 42 44.4 4222244226 4 2224243 2222242

4 .2 2.4.4. 44.4. 42 2.4.4. 48422422: 2442 22222 22242 4 2224243 2222242

4 .4 .2 42.4 42.4 42 22.4 82 4244222 42242244244442 4 2224243 2222242

4 .4 .4 .2 44.4. 44.4. 42 44.4. 4284 422224242 4 2224243 2222242

4 .4 .2 44.4 24.4 42 24.4 224243 448240 4 2224243 2222242

4842244. 228:2 44.322 4244.452
42.444.24.242

 

 

 

.4 3322 849:8 .2224 525.232.2224 425224—8423 82:22.22..— .3 own-.2 .2224 H4493. .2. 054,—.

15

to 12th rib fat thickness and weaning weight to cutability, found mean genetic correlations
of 0.37 and -0.11 and phenotypic correlations of 0.21 and -.019, respectively. These
correlations indicate that genetic improvement should occur when producers use breeding
values to make their selections, but there is an antagonistic phenotypic response on these
traits from environmental inﬂuences. Woodward et al. (1992) analyzed 13,670 Simmental
records ﬁom 1971-1988 with a multiple trait sire model and reported a genetic and
phenotypic correlation of 0.16, and 0.02, respectively, between weaning weight and
marbling. Seven hundred thirty-six Hereford bulls analyzed by Veseth et al. (1993) with
Henderson’s Method 3 (1953) produced genetic and phenotypic correlations for weaning
weight to marbling of 0.81 and 0.16, respectively. Marshall (1994) reported respective
genetic and phenotypic correlations for weaning weight to marbling of 0.39 and 0.08.
Meanwhile, Arnold et al. (1991) estimated a genetic correlation of -0.01, while Koots et
al. (1994b) reported a mean genetic correlation of -0.17 and a phenotypic correlation of
-0.04 for weaning weight and marbling. These citings indicate that a medium to high
response in carcass weight, longissimus muscle area, and other grth traits would occur
when selecting for increased weaning weights. Nonetheless, only a moderate to low
response in 12th rib fat thickness or marbling score would result when selecting for

weaning weight.

2.2.3 Yearling Weight
Yearling weight of a calf is a primary estimate of a calf‘s post-weaning growth.

Arnold et al. (1991), Johnston et al. (1992), and de Rose ( 1992) reported respective

l6

heritability estimates of 0.14, 0.16, and 0.30 (Table l). Nunez-Dominguez et al. (1993)
used a multiple trait sire model on the Charolais sired progeny on the US. Meat Animal
Research Center’s Germ Plasm Evaluation project to report a yearling weight heritability
of 0.42. Koots et al. (1994a) reported a mean heritability value of 0.35 from 154
references that had a standard deviation of 0.22. The values indicate that moderate
genetic gain can occur when selecting for increased yearling weights.

The average literature genetic correlation for each growth trait combination was
0.48, 0.42, 0.49, and 0.63, respectively, for yearling weight correlated to birth weight,
weaning weight, post-weaning average daily gain, and end of test weight. Koots et al.
(1994b) indicated a genetic correlation of -0.37 for yearling weight to marbling and a
genetic correlation of 0.32 for yearling weight to 12th rib fat thickness. Again, the
literature indicates with moderately high genetic correlations that selecting for increased
grth should result in heavier carcasses with larger longissimus muscle areas and less

marbling.

2.2.4 Post-weaning Average Daily Gain

Post-weaning average daily gain is measured from when the calf begins the post-
weaning test until the end of test weight is taken. Benyshek (1981) analyzed 8474
Hereford steers that were raised from 1960-1977 for the American Hereford Association’s
sire evaluation program with a restricted maximum likelihood (REML) algorithm to report
a post-weaning average daily gain heritability of 0.52. Koch et al. (1982) used data from

the US. Meat Animal Research Center’s Germ Plasm Evaluation project from 1970-1976

17

on 2453 steers of diverse biological types to report a post-weaning average daily gain
heritability value of 0.57. Literature values of 0.13, 0.19, and 0.47, respectively, were
reported by Arnold et al. (1991), Johnston et al. (1992), and Veseth et al. (1993) (Table
1). Koots et al. (1994a) averaged 24 heritability estimates for post-weaning average
daily gain and reported a mean of 0.29 and a standard deviation of 0.20. This mean
estimate indicates that medium genetic progress can be made for improving post-weaning
average daily gains.

A genetic correlation of 0.89 and a phenotypic correlation of 0.72 was reported by
Koch et al. (1982) between post-weaning average daily gain and carcass weight. Also,
Koch et al. reported a genetic correlation of 0.05 for post-weaning average daily gain with
12th rib fat thickness and a phenotypic correlation of 0.17 for that trait combination.
Additionally, they reported similar correlations between other growth traits with post-

weaning average daily gain (Table 2; Table 5).

2.2.5 End of Test Weight

The end of test weight is taken at the conclusion of the post-weaning test period.
A literature search for end of test weight heritability values yielded estimates of 0.52
(Benyshek, 1981) and 0.42 (Veseth et al., 1993). Koots et al. (1994a) analyzed 12
estimates which had a mean value of 0.37 with a standard deviation of 0.23 (Table 1).
These estimates indicate medium to high genetic progress can be made when selecting to

improve end of test weights.

18

Veseth et al. (1993) reported a genetic correlation of 0.28 between birth weight
and end of test weight with a phenotypic correlation of 0.24. On the same data set of
Hereford bulls, Veseth et al. determined a genetic correlation of 0.98 and a phenotypic
correlation of 0.90 between end of test weight and carcass weight. Mean genetic and
phenotypic correlations of 0.38 and 0.30, respectively, between end of test weight and
longissimus muscle area were reported by Koots et al. (1994b) as well as -0.20 and -0.22
between end of test weight and cutability. The literature review indicates higher
correlations exist between end of test weight and other grth traits that are measured as
the age of the animal increases. These are to be expected as the traits are very similar in
nature. Additionally, the mean cited genetic correlation values suggest that improvement
in end of test weights should result in moderate to high response in longissimus muscle

area but an antagonistic response in cutability.

2.2.6 Carcass Weight

A hot carcass weight is taken to assess carcass mass. Cundiﬂ‘ et al. (1971) used a
regression adjustment for age at slaughter to adjust data on 503 Angus, Hereford, and
Shorthom steers that were born from 1961-1965 and reported a carcass weight heritability
of 0.56. Koch et al. (1982) used Henderson’s Method 3 (1953) to analyze 2453 steers of
diverse biological types from the US. Meat Animal Research Center’s Germ Plasm
Evaluation project from 1970-1976 to report a carcass weight heritability of 0.43. Other
age constant heritability values for carcass weight in the literature yielded estimates of

0.24, and 0.09 from Arnold et al. (1991), and Johnston et al. (1992), respectively.

19

Additionally, age constant carcass weight heritabilities included 0.38 and 0.41 (Nunez-
Dominguez et al., 1993; Marshall, 1994) (Table 1). Wilson et al. (1993) utilized a
multiple trait REML sire model that included an adjustment for slaughter age to analyze
9448 Angus records from the American Angus Association’s carcass evaluation program
and reported a carcass weight heritability of 0.31. Koots et al. (1994a) analyzed 19 age
constant heritability estimates and reported a mean of 0.45 and a standard deviation of
0.22. These moderate and high heritability values suggest that progress can be made when
selecting for carcass weight.

A review of the literature reveals high phenotypic and genetic correlations between
carcass weight and growth traits. Marshall (1994), respectively, reported genetic and
phenotypic correlations of 0.82 and 0.62 for carcass weight and weaning weight, along
with 0.93 and 0.72 for carcass weight and yearling weight. Wilson et al. (1993), after
analyzing 10,733 Angus ﬁeld records, discovered a genetic correlation of 0.47 and a
phenotypic correlation of 0.43 between carcass weight and longissimus muscle area (Table
3; Table 6). However, Wilson et a]. (1993) found respective lower genetic correlations of
0.38 and -0.06, respectively, for carcass weight to 12th rib fat thickness and carcass
weight to marbling. These references indicate that selection for increased carcass weight
will result in faster growing, heavier muscled cattle that are fatter and have lower marbling
scores.

Benyshek ( 1981) reported that carcass weight constant heritability values for
carcass traits were lower than when both live weight and age were held constant. He

additionally indicated that once carcass traits were adjusted for age, additional adjustment

20

for carcass weight had little eﬂ‘ect on heritability. Age, carcass weight, and ﬁnish
adjustments remove the eﬂ'ects of size (scale), growth, and maturity, respectively.
Therefore, traits adjusted for these different end points are biologically different (Koots et

al., 1994a).

2.2.7 Longissimus Muscle Area

The longissimus muscle area is a measurement of the area of the longissimus dorsi
muscle measured between the 12th and 13th rib. The longissimus muscle area is used as
an indicator of total muscle because it is easy to measure and is one of the more valuable
cuts of meat from the carcass. Previously reported heritability values on an age constant
basis for longissimus muscle area include 0.41, 0.40, 0.56, and 0.46, ﬁ'om Cundiﬂ‘ et al.
(1971), Benyshek (1981), Koch et al. (1982), and Arnold et al. (1991), respectively (Table
1). Van Vleck et al. (1992) utilized records on 1350 cattle from the US. Meat Animal
Research Center’s Germ Plasm Evaluation program and analyzed the data with a single
trait animal model that accounted for slaughter age to report a longissimus muscle area
heritability of 0.60. Additional age constant heritability values include 0.51, 0.32, 0.37
(Veseth et al., 1993; Wilson et al., 1993; Marshall, 1994). In 1994a, Koots et al.
summarized 16 references and found a mean age constant heritability value of 0.43 with a
standard deviation of 0.21. Additionally, Koots et al. (1994a) located 15 references which
had a mean carcass weight constant heritability value of 0.41 with a standard deviation of
0.15. Other cited carcass weight constant heritability values for longissimus muscle area

include 0.40 (Brackelsberg et al., 1971), and 0.32 (Cundiﬂ‘ et al., 1971). Brackelsberg et

21

al. (1971) used a single trait sire model that adjusted for carcass weight on Angus and
Hereford records that spanned a four year period to analyze their data.

Koots et al. (1994b) reported mean genetic correlations of 0.40 for longissimus
muscle area to weaning weight and 0.38 for longissimus muscle area and end of test
weight. Koch et al. (1982) published a genetic correlation of -0. 14 for longissimus muscle
area to marbling score, while also showing a genetic correlation for longissimus muscle
area to 12th rib fat thickness of -0.44. Higher cutability, faster growing, lower marbled

cattle should result when selection for improved longissimus muscle area occurs.

2.2.8 Marbling

Marbling scores indicate the relative amount of intramuscular fat located within the
longissimus dorsi. The literature reviewed had marbling scores associated with numerical
values, with the greatest amount of marbling corresponding to the highest numeric value.
A review of the literature revealed heritability estimates on an age constant basis of 0.31,
0.47, 0.40, 0.35, and 0.45 (Cundiff et al., 1971; Benyshek, 1981; Koch et al., 1982;
Arnold et al., 1991; and Van Vleck et al., 1992). Barkhouse (1993) obtained records on
1432 cattle of varying breed groups ﬁom the US. Meat Animal Research Center’s Germ
Plasm project and analyzed the data with a multiple trait sire model and restricted
maximum likelihood with an age at slaughter covariate to yield a heritability estimate of
0.40. Additional age constant estimates include 0.23, 0.31, 0.26, and 0.35 ﬁ'om
Woodward et al. (1992), Veseth et al. (1993), Wilson et al. (1993), and Marshall (1994),

respectively (Table l). A mean age constant heritability estimate of 0.37 was observed

22

ﬁom 12 estimates from Koots et al. (1994a) that had a standard deviation of 0.18. Also,
heritability estimates on a carcass weight constant basis include 0.73 (Brackelsberg et al.,
1971), 0.33 (Cundiﬁ‘ et al., 1971), and 0.28 (Veseth et al., 1993). With a carcass weight
constant, Koots et al. (1994a) detected four references which had a mean heritability value
of 0.37 with a standard deviation of 0.03. These medium and high heritability estimates
show improvement in marbling scores can be achieved through selection.

A review of the literature (Arnold et al., 1991; Woodward et al., 1992; and
Marshall, 1994) exhibits relatively low genetic and phenotypic correlations between
marbling and growth traits. Nevertheless, Marshall (1994) shows a mean genetic
correlation of 0.37 between marbling score and 12th rib fat thickness from four estimates,
while ﬁve papers reviewed by Koots et al. (1994b) indicated a mean genetic correlation
between marbling score and cutability of -0.54. Therefore, selection for improved
marbling will result in little change in grth traits, but correlated responses would
indicate an increase in 12th rib fat thickness with decreasing longissimus muscle area and

cutability.

2.2.9 Twelfth Rib Fat Thickness

Fat thickness measured between the 12th and 13th rib, 3/4 of the lateral length of
the longissimus muscle measured perpendicular from the split chine bone, gives an
estimate of the amount of external fat that the carcass possesses. Due to the economic
inefﬁciencies associated with fat accretion, its importance in selection programs is of

concern. Cundiff et al. (1971), Benyshek (1981), Koch et al. (1982), and Arnold et al.

23

(1991), reported age constant heritability estimates of 0.50, 0.52, 0.41, and 0.49,
respectively (Table 1). Likewise, age constant heritability estimates of 0.26 (Wilson et al.,
1993), and 0.44 (Marshall, 1994) were determined, while Koots et al. (1994a) reported 26
estimates which had a mean of 0.43 and a standard deviation of 0.18 on an age constant
basis. Other references in the literature on a carcass weight constant basis include 0.43
(Brackelsberg et al., 1971), and 0.53 (Cundiff et al., 1971). Fifteen references cited by
Koots et al. (1994a) yielded a mean carcass weight constant heritability value for 12th rib
fat thickness of 0.44 with a standard deviation of 0.15. These predominantly high
heritability estimates suggest rapid changes in 12th rib fat thickness in beef cattle can be
made.

Koots et al. (1994b) reported 10 genetic correlation estimates between 12th rib fat
thickness and weaning weight that had a mean of 0.07, along with four genetic
correlations between 12th rib fat thickness and end of test weight with a mean of 0.02.
Additionally, Wilson et al. (1993) indicated a genetic correlation between 12th rib fat
thickness and carcass weight of 0.38, while Koch et al. (1982) reported a genetic
correlation of 0.16 between 12th rib fat thickness and marbling along with a genetic
correlation of -0.44 between 12th rib fat thickness and longissimus muscle area. The
genetic correlations reported in Table 2 indicate selection against 12th rib fat thickness
will increase longissimus muscle area and cutability but decrease grth and marbling

scores.

24

2.2.10 Carcass Grade

The Canadian Meat Council’s carcass grades are composed of two factors, meat
quality and carcass meat yield. Quality factors are composed of 12th rib fat thickness,
marbling, fat and meat color, maturity, meat ﬁrmness, adequate muscle thickness, pizzle
eye size, and crest development (an increase in the mass of various neck muscles).
Carcass meat yield is predicted ﬁom 12th rib fat thickness and longissimus muscle area
measurements (Jones, 1993). Young carcasses are broken down into two categories, A or
B. If a carcass has at least Traces marbling, at least 4 mm of 12th rib fat thickness, good
muscling, bright red meat color, and white fat, then it qualiﬁes for the A grade. The A
grade then is broken down due to different carcass yield percentages that follow: A1
259% carcass yield (4-10 mm 12th rib fat thickness); A2 is from 54-58% carcass yield
(10-15 mm 12th rib fat thickness); and A3 is 353% (>15 mm 12th rib fat thickness). A
carcass that has less than 4 mm of 12th rib fat thickness or less than Traces marbling
receives a BI grade, while carcasses that have yellow fat, poor (light) muscling, or dark
colored meat will receive a B2, B3, or B4 carcass grade, respectively. A review of the
literature yielded no heritability estimates or correlation values for carcass grade under the

Canadian grading system.

2.2.11 Cutability
Percent cutability is an estimate of the lean primal cuts from the carcass. Hot

carcass weight, longissimus muscle area, and 12th rib fat thickness are the measurements

25

which compromise cutability. The formula for percent cutability from the Lacombe

Research Station is as follows:

percent cutability = 53 - 7 (12th rib fat thickness) + 0.7 (longissimus muscle area).

Cundiﬂ‘ et al. (1971), Benyshek (1981), Woodward et al. (1992), and Marshall (1994)
reported respective age constant heritability estimates of 0.28, 0.49, 0.18, and 0.36 (Table
1). Koots et al. (1994a) summarized 12 age constant heritability estimates to have a
mean and standard deviation of 0.41 and 0.14, respectively. These estimates indicate that
moderate to high genetic progress can be made if selection is for cutability.

In the literature, there appears to be low genetic correlations between cutability
and growth traits (Table 4) (Woodward et al., 1992; Marshall, 1994; Koots et al., 1994b).
Koots et al. (1994b) reported a mean genetic correlation of 0.12 between cutability and
carcass weight ﬁ'om three literature references, while Marshall (1994) averaged two
literature references and indicated a negative genetic correlation for this same trait
combination of -0. l 1. Additionally, Koots et al. (1994b) described a mean genetic
correlation of 0.26 between longissimus muscle area and cutability, and -0.33 between
12th rib fat thickness and cutability. Incidentally, Koots et al. (1994b) reported a negative

mean genetic correlation of -0.54 ﬁ'om ﬁve sources between cutability and marbling score.

26

2.3 Parameter Estimation

2.3.1 Model Specification
The evolution of sire evaluation yields a dependency upon certain criteria
(Henderson, 1973, 1974):
1. The predictor has the same expectation as the unknown variable that is to
be predicted.
2. Minimization of the variance of the error of prediction in the class of linear
unbiased predictors.
3. Maximization of the correlation between the predictor and the predictand
in the class of linear unbiased predictors.
4. When the distribution is multinomial normal:
a. yields the maximum likelihood and the best linear unbiased
estimator of the conditional mean of the predictand.
b. in the class of linear, unbiased predictors, maximizes the
probability of a correct pairwise ranking.
The Best Linear Unbiased Prediction (BLUP) developed by Henderson (1953) follows
these criterion. However, Henderson (1975) recognized possible errors ﬁ'om model
misspeciﬁcation with BLUP. Ignoring relevant ﬁxed eﬁ‘ects yielded biased estimators.
Also, the inclusion of irrelevant factors increased the sampling variance, while if random
factors were excluded, even if relevant, the estimator and predictor would remain unbiased
even though an increase in the sampling variance results.
Pollak and Quaas (1980) make several points comparing an “animal model” to a
sire model. The equation for a record contains a term for the breeding value of the animal
making the record. This allows for the possibility that sires and(or) dams may have

records. It also means that evaluations will be obtained for animals that are not sires (or

dams). A less desirable consequence is that the number of random elements to be

27

predicted, at least one for each record, becomes exceedingly large. Other basic discussion
of these two types of models include that in a sire model, the animal vector contains only
the additive genetic effects ﬁ'om the male parents. One basic concept of a sire model is
that the sires are randomly bred to dams and that the dams have only one progeny. The
use of an animal model removes the potential bias ﬁom these assumptions as it models all
animals in the population, including those without records, and connects the animals
through the additive genetic relationship matrix.
The basic form of mixed linear models with one random factor is as follows:
y = Xb + Zu + e

where: y is an N X] vector of observations,

b is a p x 1 vector of ﬁxed effects associated with y,

u is a q x 1 vector of random effects associated with y,

X is a known incidence matrix of order N x p that relates elements of b to

elements of y,
2 is a known incidence matrix of order N x q that relates elements of u to
elements of y, and

e is an N x 1 vector of residual effects.
Additional attributes of the general form of mixed linear models include the expectations
of the random variable which include:

130’) = Xb.
E(u) = 0, and

E(e) = 0.

28

The (co)variance structure is:

y ZGZ'+R ZG R
V u = GZ' G 0 .
e R 0 R

2.3.2 Single and Multiple Trait Mixed Model Methodology

The single trait mixed model equations described by Henderson’s (1953) BLUP
methodology assume that all correlations between traits are zero. An extension of
Henderson’s BLUP is accomplished through multiple trait analysis that improves the
accuracy of genetic evaluation, especially of traits that have been selected upon or are
lowly heritable. In animal breeding, most populations have been selected upon, and
generally for more than one trait. An example of sequential selection occurs when
observations for one trait are used to cull animals, and the selected group is then measured
for subsequent traits. Single trait mixed mode] evaluations would therefore be potentially
biased by selection for the ﬁrst trait.

Multiple trait analysis eliminates the bias due to sequential selection and also
selection on correlated traits (Pollak et al., 1984; Walter and Mao, 1983). Also,
depending upon the genetic and error correlations used, a reduction in prediction error
variance and therefore an increase in accuracy occurs through the use of multiple trait
analysis (Schaeffer, 1984). Additionally, Schaeﬂ‘er (1984) points out that multiple trait

analysis allows all animals to be evaluated for every trait even without individual

29

observations for each trait because of the non-zero genetic and residual covariances
among the traits that are included in the analysis.

An advantage of single trait analysis is that there are less equations to be solved
through the iteration process (Banks, 1986; Nwerume, 1994). Convergence can slow as
the number of traits increases, and the complexity of multiple trait models increases rapidly

past two traits.

2.3.3 (Co)variance Estimation

Due to the computational difﬁculty of ﬁtting the expectations and reductions for
large data sets, other techniques such as the method of maximum likelihood are preferred.
Hartley and Rao (1967) present a maximum likelihood (NIL) method that is applied to the
general mixed model. Patterson and Thompson (1971) published a restricted maximum
likelihood method (REML) of estimating intra-block and inter-block weights in the
analysis of incomplete block designs with block sizes that are not necessarily equal.
Another view of the problem is estimating constants and components of variance from
data arranged in a general two-way classiﬁcation when the effects of one classiﬁcation are
regarded as ﬁxed and the eﬂ‘ects of the second classiﬁcation are regarded as random. The
method they described takes the expectations over a conditional distribution with the
treatment effects ﬁxed at their estimated values. Their method consists of maximizing the
likelihood, not of all the data, but of a set of selected error contrasts, using iterative

techniques. Additionally, Harville (197 7) describes techniques such for a given statistical

30

model, when the estimating 9 parameters, and with an assumed data distribution, the
likelihood function L(9) can be equated. In Harville’s review of ML and REML, he notes
Patterson and Thompson’s computationally feasible REML method takes into account the
loss of degrees of freedom resulting ﬁom the estimation of the ﬁxed eﬂ‘ects (Cunningham,
1989)

A derivative-ﬂee algorithm for use during the restricted maximum likelihood (DF-
REML) variance component estimation was presented by Graser et al. (1987). This
method avoids explicit evaluation of ﬁrst derivatives and does not require matrix
inversion. As a result, one round of the method involves computing the determinant of the
coeﬂicient matrix of the mixed model equations, which uses a one-dimensional search
involving the variant part of the log likelihood to ﬁnd the maximum of this function
(Saama, 1992). The derivative-free multivariate REML algorithms are computationally
expensive, especially if the likelihood function contains many parameters to be estimated.
Due to this, other algorithms have been developed which utilize ﬁrst and second
derivatives of the likelihood function.

A REML algorithm which uses considerably less computer time, but gives almost
identical parameter estimates as DF-REML, is named AI-REML (Madsen et al., 1994).
AI-REML uses the average of the observed and expected information as the information
matrix. The matrix of second derivatives is called the observed information matrix.
Expectations of this matrix is the Fisher information matrix. REML algorithms which
utilize observed or expected information will lead to either the Newton-Raphson or the

Fisher-scoring algorithm, respectively. The terms that are involved in computing either

31

respective information matrix are computationally diﬁicult. Johnson and Thompson
(1995) showed the average of the observed and expected information matrix was
considerably easier to compute than either the observed or expected information matrix
due to the cancellation of terms. Therefore, a compromise between the Newton-Raphson

and the Fisher-scoring algorithms is AI-REML (Madsen et al., 1994).

2.3.4 Genetic Parameter Estimation

2.3.4.1 Heritability Estimators

Falconer (1960) indicates that the variation of a record can be reduced into three
categories: additive variance, dominance variance, interaction (epistatic) variance.
Additive variance is the variance of breeding values or the variance that can be passed on
to offspring. Dominance variance is variation of a record that is due to dominant alleles,
or genes that are more “robust” during segregation. Interaction variance is variance due to
genes that interact during segregation and expression. An example of an interaction can
be from additive by dominance effects. Generally, the eﬁ‘ects of dominance variation and
interaction variation are considered to be low and are not as readily assessed through
observations made on the population, so in practice, the most important partition is due to
additive variation. Falconer (1960) noted that the single most important ﬁrnction of the
heritability in the genetic study of metric characters is its predictive role, expressing the

reliability of the phenotypic value as a guide to the breeding value. Phenotypic values of

32

individuals can be directly measured, but it is the breeding value that most often
determines their inﬂuence on the next generation. Therefore if the breeder or
experimenter chooses individuals to be parents according to their phenotypic values, his
success in changing the characteristics of the population can be predicted only from a
knowledge of the degree of correspondence between phenotypic values and breeding
values. This degree of correspondence is measured by the heritability.

Falconer (1960) states that the half-sib correlation and the regression of offspring
on father is a reliable heritability estimate for attempting to reduce sampling error and also
environmental sources of covariance that cannot be statistically overcome. The formula
for heritability estimates for trait i from paternal half-sibs is as follows:

A

2
A2 40

i - '2 "2
0"I+(r.l

A

where: h. 2 is estimated heritability,

is the estimated sire variance component,

is the estimated error variance component.

The denominator is the phenotypic variance adjusted for ﬁxed effects which were

included in the model.

3.3.4.2 Correlation Estimators
Falconer (1960) described the pleiotropic action of genes, which is the property of

a gene to affect two or more characters when segregating, to cause simultaneous variation

33

in two or more traits. The correlation of breeding values is the genetic correlation, while
the environmental correlation is the correlation of environmental deviations together with
non-additive genetic deviations. A correlation is the ratio of the appropriate covariance to
the product of the two standard deviations. The genetic correlation between two traits
can be affected by selection if selection has been placed on the parents; and as the genetic
correlation increases, the bias increases in a likewise direction of the estimated correlation

with intense selection (Van Vleck, 1968).

3. MATERIALS AND METHODS

3.1 Description of the Data and Data Edits

Data for this project were provided by the Canadian Charolais Association’s
Conception to Consumer Program. The Conception to Consumer Program is a sire
evaluation program which gives members of the Canadian Charolais Association an
opportunity to sponsor bulls to be evaluated through a progeny test for growth and
carcass characteristics. The program was initiated in 1968.

Annually, bulls are nominated by the producers for the program and are randomly
bred to mature cows in cooperator herds located in Alberta, British Columbia, or
Saskatchewan. The breed makeup of the cowherds consisted of varying breeds and breed
combinations and were intended to be representative of 90% of the Canadian national beef
herd. No virgin heifers were used as dams in the program. Birth weight records were
collected by the cooperator herd operators, while the weaning weights were collected
under the supervision of the Canadian Charolais Association.

In the fall (late October or early November), calves were weaned at an average age
of 215 days and transported to the Cattleland Feedlot, Strathmore, Alberta. At the
feedlot, cattle were managed under typical commercial practices and were under the
supervision of the Canadian Charolais Association. Upon arrival at the feedlot, cattle
were processed (given typical medication and induction treatments), sorted by sex and

weight, and then were fed an adaptation ration for an average of 48 days. Following the

34

35

adaptation period, the steers and heifers were placed on a bulk cereal-silage based diet for
an average of 115 days until the end of test date. Following the end of test, cattle
remained on the same ration until they were visually appraised to have reached the Al or
A2 carcass grade, at which time they were slaughtered. On average, the cattle were
slaughtered 41 days after they came off test. The carcass grades are ﬁom the Canadian
Meat Council and correspond to the respective twelfth rib backfat thickness of 4-10 and
10-15 mm for the Al and A2 carcass grades.

After being delivered to the packing plant, all calves were slaughtered and “blue
tagged,” utilizing the program provided by Agriculture Canada. The meat graders collect
the carcass information, which included hot carcass weight, 12th rib backfat thickness,
longissimus muscle area, marbling score, and carcass grade. The backfat thickness was
measured in tenths of inches, between the 12th and 13th rib, 3/4 of the lateral length of the
longissimus muscle measured from the split chine bone. The longissimus muscle area
measurement was taken in square inches and measured to the nearest tenth. Marbling
scores were also taken at the 12th and 13th rib section, and were classiﬁed using a nine
point scale with a lower numeric number equating to more marbling. The numeric scores
were: 1 = very abundant; 2 = abundant; 3 = moderately abundant; 4 = slightly abundant; 5
= moderate; 6 = modest; 7 = small; 8 = slight; and 9 = traces. The carcass meat quality
and meat yield attributes were evaluated to place the carcass into one of seven carcass
grades: Al, A2, A3, A4, B1, B2, and B3. These carcass grades were then associated
with a numeric value to make the analysis possible. The association is: A1 = 1; A2 = 2;

A3 = 3; A4 = 4; B1 = 5; B2 = 6; B3 = 7. Percent cutability was estimated as the percent

36

of lean primal cuts. The lean primal cuts included closely trimmed boneless chuck, rib,
loin, and round cuts. The formula for percent cutability from the Lacombe Research
Station is as follows:

Percent cutability = 53 - 7 (12th rib backfat thickness)

+ 0.7 (longissimus muscle area)

Contemporary groups were deﬁned as animals in the same calving year, breeder
herd, sex of calf, breed of dam group, and weaning group. The breed of dam was a visual
appraisal given by the cooperating herd operator and is consistent only within herds.

The data set initially contained 5497 records from 1975 through 1993, but was
edited to remove single record contemporary groups, gross recording errors, and all twin
records (Table 8). There were 9 single record contemporary groups, 2 gross recording
errors (>7 standard deviations away from the trait mean), and 85 twin records. No
reference sires were used during the ﬁrst 3 years of the program, but all 172 sires included
in the program were connected through the use of pedigree additive relationship
information on the sires (Henderson, 1974).

After edits, the resulting data set contained 5401 records which contained 368
contemporary groups, ranging in size from 2 to 141 calves. Fifty-three percent of the
5401 records were ﬁ'om steers. Forty-seven percent of the records were from heifers.
Table 9 indicates the reduction in records by sex at different dates that traits were

measured. Of the heifer records, there was a decrease in the number of records ﬁ‘om

37

Table 8. Data editing criteria and number of records deleted.

 

 

Editing Criteria Number of Records Deleted
Twin Records 85
Single Record Contemporary Groups 9
Gross Recording Errors 2

 

Total Records Edited 96

 

38

Table 9. Decrease of volume of records by sex.

 

 

Trait Dat'e Male Female
Calving Year 2881 2520
Weaning Date 2660 2360
Start of Test Date 2383 1161

Slaughter Date 2162 1022

 

39

weaning (2360 records) to the start of the test period (1161 records). The loss of 1199
heifer records was a 50.8 percent decrease that indicates that selection occurred upon the
heifer population from the date these two traits were measured.

The traits in the analysis include: birth weight (BW); 200 day age adjusted
weaning weight (WW); 365 day age adjusted yearling weight (YW); post-weaning
average daily gain on test (ADG); end of test period weight (EOTWT); hot carcass weight
(CARCWT); marbling score (MARB); longissimus muscle area (LMA); 12th rib fat

thickness (FAT); carcass grade (CARCGR); and cutability (CUT).

3.2 Estimation of Genetic Parameters
Following the edits for single record contemporary groups, gross recording errors,
and all twin records, the number of records, means, and standard deviations were

calculated and are presented in Table 10.

3.2.1 Model 1: Single Trait Mixed Model

Initial variance component estimates were estimated through the use of the PROC
VARCOMP procedure from SAS Institute (1990). In the initial analysis, a simple model
was used due to computational limitations. The model equated the dependent variable to
contemporary group effect and a random sire effect. No sire relationship information was

included. The model used included:

ﬁzﬂh+zm+q

40

where:

i = l, 2, 3, . . ., 11 which corresponds to birth weight, adjusted weaning weight,
adjusted yearling weight, post-weaning average daily gain, end of test
weight, carcass weight, marbling score, longissimus muscle area, 12th rib
fat thickness, carcass grade, and cutability,

y; is an observation vector of the 1"" trait,

b; is vector of the ﬁxed effect for contemporary group for the i”I trait,

u.- is a vector of random genetic sire effects for i” trait,

X.- is an incidence vector associating ﬁxed eﬁ‘ects with the corresponding
records in y,

Z.- is an incidence vector associating random sire genetic effects with the
corresponding records in y, and

e,- is the random error effect for y.

The expectations for the model include:

ELy] = Xb
E[ozs] = 14024
E[e] = 0.

where 024 is deﬁned as the additive genetic variation. The variance-covariance matrix

41

associated with the random eﬂ‘ects in the model is:

where:

and where a; is the variance associated with the sire effect, and of, is the residual

variance.

3.2.2 Model 2: Single Trait Mixed Model

Once the initial variance components were estimated using the simple model, a
more complex single trait model was run using the multiple trait, average-information
restricted maximum likelihood algorithm (AI-REML) of DMU (DMU-AI) (Jensen and
Madsen, 1993) on a IBM RISC System 6000 computer. The prior variance estimates for

Model 2 were taken from the VARCOMP procedure results. The models used included:

J" = Xrﬂi +24”: +3.:

Table 10. Number of records after edits, means, and standard deviations.

 

 

Trait Number of records Mean Standard Deviation
Birth Weight 5223 42.24 kg 6.44 kg
Adjusted Weaning Weight 4801 247.78 kg 35.24 kg
Adjusted Yearling Weight 3330 466.19 kg 56.24 kg
Post-weaning Average Daily Gain 33 30 1.47 kg/day 0.025 kg/day
End Of Test Weight 3441 482.07 kg 53.03 kg
Carcass Weight 3176 308.16 kg 33.50 kg
Marbling Score 3184 6.93 units 0.75 units
Longissimus Muscle Area 3145 81.61 cm2 9.81 cm2
12th Rib Fat Thickness 3143 0.889 cm 0.305 cm
Carcass Grade 3166 1.32 units 0.57 units
Cutability 3143 58.94 % 1.48 %

 

43

where:

i = 1, 2, 3, . . ., 11 which corresponds to birth weight, weaning weight, yearling
weight, post-weaning average daily gain, end of test weight, carcass
weight, marbling score, longissimus muscle area, 12th rib fat thickness,
carcass grade, and cutability,

y.- is an observation vector of the 1"" trait,

,6.- is matrix of the ﬁxed eﬂ’ects for 1"" trait,

u.- is a vector of random genetic sire effects for 1"” trait,

X.- is an incidence matrix associating fixed effects with the corresponding
records in y,

Z.- is an incidence matrix associating random sire genetic effects with the
corresponding records in y, and

e,- is the random error eﬁ‘ect for y.

The expectations for the model include:
Efy] = Xb
E[ozs] = 3102,.
E[e] = 0.

where 02,. is deﬁned as the additive genetic variation. The variance-covariance matrix

44

associated with the random effects in the model is:

where:

and where a; is the variance associated with the sire effect, and of, is the residual
variance. Also, G = G, ®A, and R = R, 81 , as A is the numerator of the additive

genetic relationship matrix among the 172 sires with ® denoting the direct product
operator.

With these results, the mixed-model equations are:
X'R"X X'R"Z ,B _ X'R"y
Z'R"Z Z'R"Z + G" u " Z'R“ y

where: G " = G,” ®A" , and R" = R;' 8) I . The A‘1 elements were established using

methods described by Henderson (1976) and Quaas (1976).
The ﬁxed eﬂ‘ects in the model included 11 contemporary group and also one

covariate, which was age at slaughter. Age at slaughter was included as a covariate in the

45

model for carcass weight, marbling score, longissimus muscle area, 12th rib fat thickness,

carcass grade, and cutability because the cattle were slaughtered at diﬂ‘erent ages to

achieve a visual A1 or A2 carcass grade, and also to study differences in the carcass traits

on a grth rate constant basis (Benyshek, 1981).

The algorithm used in the analysis is as follows (Johnson and Thompson, 1995;

Madsen et al., 1994):

where:

I

5(9) = F'PF = F'R" - (C"W'F) WR”F

[4(6) = average information matrix,

F = a matrix with the dimension of (number of observations by number of

 

5V
RV.
{1’}

elements in 6), with the j” column 1} equal to the vector

P = V” —V"X(X'V"X)"X'V-',

R = residual (co)variance matrix,

C I = inverse of the coeﬁicient matrix of the mixed model equations,
W = total (ﬁxed and random eﬂ‘ects) design matrix,

V = variance matrix of y = ZGZ’+R,

X = design matrix corresponding to the ﬁxed effects,

y = vector of observations,

G = variance of u,

6 = scalar variance component,
Z = design matrix corresponding to the random effects.

Therefore, for a column in F corresponding to an element in G,‘ , in such that Go, is a
p, x p, (co)variance matrix of the traits in the 1"" random effect, Madsen et al. (1994)
indicate:
6V .
f 64131.} : [—]Py = Zr (D'chr-,l)®l r
m .. I, }«
where: D}; is a symmetric p; x p, indicator matrix containing ones in positions
corresponding to the j,k" parameter in G," and zeros elsewhere,
6,, 11*} is the corresponding element in 6, and
:3, is the vector of all solutions for the 1"" random eﬂ‘ect.

To estimate the columns in F that correspond to the parameters in R. the algorithm is:

r()[..i]w(x»>

The convergence criterion used in all analyses was set at a level where the (co)variances

changed less than] x 10‘8 standard deviation units per round of iteration. After the initial
run of the model reached convergence, the (co)variance components were used as starting
values in a subsequent cold restart. After all restart runs were completed, the respective
(co)variance results were averaged and the mean estimates and standard deviations were
reported. Heritabilities and genetic, phenotypic, and environmental correlations were
estimated ﬁom the paternal half-sib variances and covariances ﬁ'om the ﬁve trait analyses.

Heritability (the proportion of the phenotypic variance which is explained by the additive

47

genetic variance) was estimated from intraclass correlations of paternal half-sibs (Falconer,
1960):

h2 :40: lai,

where a: = of + of. Dickerson (1958) noted that this estimate of heritability may be

upwardly biased due to epistatic effects, such as any genotype by environment

interactions.

3.2.3 Model 3: Five Trait Mixed Model

A third analysis was conducted using the results from the second analysis with
single trait procedures as prior variance component values. In the third analysis, a ﬁve
trait mixed model was employed in which adjusted weaning weight (WW) was included in
each 5 trait combination. This model was used to explore the hypothesis of selection bias
in estimation of genetic parameters due to the relative increase in the loss of female post-

weaning records (Henderson and Quaas, 1976; Pollak and Quaas, 1980). The model used

48

 

 

 

 

 

 

 

 

 

 

in this analysis is as follows:
Ty , ”X 0 0 0 o Tb ‘
W W W
y, 0 X, 0 0 0 b,
y, = 0 0 X,. 0 0 b,.
y.” 0 0 0 X,. 0 b,.
' 0 0 0 0 x... b...
L3”?” 2 2 3 2L 4 2
”Z". 0 0 0 0 "aw ”cw“
0 Z, 0 0 0 u, e,
+ 0 0 Z,. 0 0 u,. + e,.
0 0 0 Z,. 0 u,. e,.
_0 0 0 0 Z,..__u,- 3 _e,...
where:

i = trait 2, 3, 4, or 5, in addition to i¢i'¢i"¢i"'¢weaning weight, all trait
combinations were computed so as to ﬁll a correlation matrix, while
still having weaning weight remain in every multivariate model,

y is an observation vector of i, and weaning weight,

b is a vector of the ﬁxed effects for i, and weaning weight, the ﬁxed effects
in the model included u contemporary group and also one
covariate, which was age at slaughter. Age at slaughter was
included as a covariate in the model for carcass weight, marbling
score, longissimus muscle area, 12th rib fat thickness, carcass

grade, and cutability

49

u is a vector of random genetic sire eﬂ‘ects for i, and weaning weight,

X is an incidence matrix associating ﬁxed effects with the corresponding
records in y,

Z is an incidence matrix associating random sire genetic effects with the
corresponding records in y, and,

e is the random error effect for i, and weaning weight.

The expectations for the model include:

E[y] = Xb
Elo’s] = 1442.
E[e] = 0
where 62A is deﬁned as the additive genetic variation. The variance-covariance matrix

associated with the random effects in the model is:

at] = [0,: 1.1]

50

where:

 

 

' 2
06,, OWWG, arr/n10, wage, Orr/mo, T
2
00, 00,0, 06,6, 0,6,-
_ 2
Go — 06,. 00,0“. 00,6,.
2
00,. 06,02
- 2
Symetrrc 04,. J
_ 2 2
04,, aRnRi aRnRr' aRnRr‘ aRnRr
2
0R. 014R. Uaa- 08R.-
__ 2
R, — 04,. “12.-R. 0 R411.-
2
UR” aRrRr-
. 2
Symetrrc 0,,”

 

 

and where 0?, are variances associated with the sire eﬁ‘ects, and of, are residual
variances. In addition, 04,4, are covariances associated with sire eﬂ‘ects, and a”, are
covariances analogous to the dependent traits analyzed. Also, G = G, 69 A , and

R = R, ® I , as A is the numerator of the additive genetic relationship matrix among the

172 sires with ® denoting the direct product operator.

With these results, the mixed-model equations are:
X'R"X X'R"Z ,B _ X'R“y
Z'R"Z Z'R"Z + G" u Z'R"y

where: G"' =G,," ®A", and R" = R;’ ®I.

and Quaas (1976).
(N ewton-Fisher REML Estimation) utilizing the multivariate mixed model package DMU

(Jensen and Madsen, 1993). The algorithm used in the analysis is as follows (Johnson and

51

The A“ elements were established using methods described by Henderson (1976)

Thompson, 1995; Madsen et al., 1994):

where:

I

1,,(0) = F'PF = F'R" - (C"W'F) WR"F

IA( 6) = average information matrix,

F = a matrix with the dimension of (number of observations by number of

 

6V
Py.
{1'}

elements in 6), with the j"I column f5 equal to the vector

P = V" — V"X(X'V-'X)"X'V",

R = residual (co)variance matrix,

C I = inverse of the coefﬁcient matrix of the mixed model equations,
W = total (ﬁxed and random eﬂ‘ects) design matrix,

V = variance matrix of y = ZGZ’+R,

X = design matrix corresponding to the ﬁxed effects,

y = vector of observations,

G = variance of u,

The (co)variance components were estimated using AJ-REML

52

6 = scalar variance component,
Z = design matrix corresponding to the random eﬁ‘ects.

Therefore, for a column in F corresponding to an element in Go, , in such that Go, is a

pr x p; (co)variance matrix of the traits in the i" random effect, Madsen et al. (1994)

indicate:

 

f(6,,,,,) = [55:4 J1» = Z,[(D,,G;')® 111?,

where: Dy, is a symmetric p; x p; indicator matrix containing ones in positions

corresponding to the j,k"‘ parameter in G,‘ and zeros elsewhere,
6,, m is the corresponding element in 6, and

ii, is the vector of all solutions for the 1"" random effect.

To estimate the columns in F that correspond to the parameters in R. the algorithm is:

“‘9er) 2 [will—J” = R4R"(y ’ XI; ’ 2")

am}

The convergence criterion used in all analyses was set at a level where the (co)variances
changed less thanl x 10" standard deviation units per round of iteration. After the initial
run of the model reached convergence, the (co)variance components were used as starting
values in a subsequent cold restart. After all restart runs were completed, the respective
(co)variance results were averaged and the mean estimates and standard deviations were
reported. Heritabilities and genetic, phenotypic, and environmental correlations were

estimated from the paternal half-sib variances and covariances from the ﬁve trait analyses.

53

Heritability (the proportion of the phenotypic variance which is explained by the additive
genetic variance) was estimated from intraclass correlations of paternal half-sibs (Falconer,
1960y

h2=403/0i,

where a; = a: + of. Dickerson (1958) noted that this estimate of heritability may be

upwardly biased due to epistatic effects, such as any genotype by environment
interactions.

The formula used for genetic correlation of trait i and i' is (Falconer, 1960):

_ 2 2
rg — 06.6,. l"06,06,. -

The formula for environmental correlation of trait i and i ' is deﬁned as (Falconer,

1960)

_ I 2 2
I; — aglgrl OEUE“ 4

The phenotypic correlation formula for trait i and i ' is (Falconer, 1960):

 

_ j 2 2 _ 2 2 2 2
rp -O-BPI’/ CHOP, — 0.610,: +0535" /J(O'G' +OEIXUGI' +015”) 4

54

An approximate method to estimate the variance of the intraclass correlation (1)

from half sibs was used (Swiger et al., 1964). The paternal half-sib correlation is

t = a: la: . The variance of t for a simple one-way classiﬁcation model is:

V(t) z- {2(n.—1)(1 — t)2[1 + (k — 1):]2} /[k2 (n.—B)(B — 1)]

where:
. = total number of animals,
B = number of sires,
k = [12.—(£11,?) / n.] / (B — 1) and,

n, = number of progeny of sire i.

Therefore, the variance of heritability is:
V(h2) = V(t) la: =16V(t)
where, a,; = 1/4, the additive relationship of paternal half sibs.

The subsequent approximate standard errors are expected to underestimate the

actual standard errors (Swiger et al., 1964).

3.2.4 Model 4: Four Trait Mixed Model
A fourth analysis was run to investigate the variation in marbling score,

longissimus muscle area, and 12th rib fat thickness while holding carcass weight constant

to observe differences in marbling score, longissimus muscle area and 12th rib fat
thickness proportions of the carcass (Cundiﬂ‘ et al., 1971). In the fourth analysis, a four
trait mixed model was used in which weaning weight was included. This model was used
to explore the hypothesis of selection bias in estimation of genetic parameters due to the

relative increase in the loss of female post-weaning records (Henderson and Quaas, 1976;

55

Pollak and Quaas, 1980). The model used in this analysis is as follows:

where:

yw
J’cum

Yam

 

score (CMARB), longissimus muscle area (CLMA), and 12th rib

 

 

J'cmr j

 

 

fat thickness (CFAT),

 

 

 

 

y is an observation vector of adjusted weaning weight (WW), marbling

56

b is a vector of the ﬁxed effects for weaning weight, marbling score,
longissimus muscle area, and 12th rib fat thickness, the ﬁxed effects
in the model included u contemporary group and also one
covariate, which was carcass weight,

n is a vector of random genetic sire effects for weaning weight, marbling
score, longissimus muscle area, and 12th rib fat thickness,

X is an incidence matrix associating the ﬁxed effects u contemporary group
and the carcass weight covariate with the corresponding records in
y,

Z is an incidence matrix associating random sire genetic effects with the
corresponding records in y, and,

e is the random error effect for weaning weight, marbling score,

longissimus muscle area, and 12th rib fat thickness.

The expectations for the model include:

E[y] = Xb
E[ozs] = 14024,
E[e] = 0

where 02,4 is deﬁned as the additive genetic variation. The variance-covariance matrix

57

associated with the random eﬁ’ects in the model is:

where:

and where 0:,
variances. In addition, 04.4,. are covariances associated with the sire effects, and a 12.12,.

are the covariances analogous to the dependent traits analyzed. Also, G = G, ® A , and

R = R, ® I , as A is the numerator of the additive genetic relationship matrix among the

mm =

 

06,, 0mm,
2
0' G,
G, =
Symetric
' 2
0R" RnR,
2
”R.
R, =
Symetric

 

G.
0R

“R444

“Rm.

I

172 sires with ® denoting the direct product operator.

are variances associated with the sire effects, and 0,2, are residual

 

 

where:

and Quaas (1976).

(N ewton-Fisher REML Estimation) utilizing the multivariate mixed model package DMU

(Jensen and Madsen, 1993). The algorithm used in the analysis is as follows (Johnson and

58

With these results, the mixed-model equations are:
X'R“X X'R“Z ,B _ X'R“y
Z'R"Z Z'R“Z +G" u Z'R"y

G" =G,"®A",and R" =R;’®I.

The A’1 elements were established using methods described by Henderson (197 6)

Thompson, 1995; Madsen et al., 1994):

where:

I

1,,(9) = F'PF = F'R" - (C"W'F) WR"F

IA (6) = average information matrix,

F = a matrix with the dimension of (number of observations by number of

 

6V
Py,
{fl

elements in 6), with the jth column f} equal to the vector

P = V" —V"X(X'V"X)"'X'V",
R = residual (co)variance matrix,

C’ = inverse of the coeﬂicient matrix of the mixed model equations,

The (co)variance components were estimated using AI-REML

59

W = total (ﬁxed and random effects) design matrix,

V = variance matrix of y = ZGZ’+R,

X = design matrix corresponding to the ﬁxed effects,
y = vector of observations,

G = variance of u,

6 = scalar variance component,

Z = design matrix corresponding to the random eﬁ’ects.

Therefore, for a column in F corresponding to an element in G,, , in such that G,, is a

p, x p; (co)variance matrix of the traits in the i‘” random effect, Madsen et al. (1994)

indicate:

A9444) = [399,—]13 2 Z‘[(DI~G:)® 1]"

rm}
where: D); is a symmetric p, x p, indicator matrix containing ones in positions
corresponding to the j,k"‘ parameter in G,‘ and zeros elsewhere,
6,1“, is the corresponding element in 6, and
ii, is the vector of all solutions for the 1"” random eﬁ‘ect.

To estimate the columns in F that correspond to the parameters in R. the algorithm is:

()[—]<)

RUM

The convergence criterion used in all analyses was set at a level where the (co)variances
changed less than] x 10'8 standard deviation units per round of iteration. After the initial
run of the model reached convergence, the (co)variance components were used as starting
values in a subsequent cold restart. Heritabilities and genetic, phenotypic and
environmental correlations were estimated from the paternal half-sib variances and
covariances from the four trait analyses. Heritability (the proportion of the phenotypic
variance which is explained by the additive genetic variance). was estimated from

intraclass correlations of paternal half-sibs (Falconer, 1960):

h2 :40: lai,

where of, = of + of. Dickerson (1958) noted that this estimate of heritability may be

upwardly biased due to epistatic effects, such as any genotype by environment
interactions.

The formula used for genetic correlation of trait i and i ' is (Falconer, 1960):

The formula for environmental correlation of trait i and i' is deﬁned as (Falconer,

1960)

61

The phenotypic correlation formula for trait i and i' is (Falconer, 1960):

 

_ 2 2 _ 2 2 2 2
r, —0',,,,, “/0503 — 00,6, +0315, “Koo, +053X06, +05.)

An approximate method to estimate the variance of the intraclass correlation (i)

from half sibs was used (Swiger et al., 1964). The paternal half-sib correlation is
t = a: /a,2, . The variance oft for a simple one-way classiﬁcation model is:

V (t) E {2(n.—1)(1 — t)2[1 + (k — 1)t]2}/[k2(n.—B)(B — 1)]
where:

. = total number of animals,

B = number of sires,
k = [n.—(2n,2) /n.] / (B — 1) and,

n; = number of progeny of sire i.

Therefore the variance of heritability is:
V(h2) = V(t) / a: =16V(t)
where, a: = 1/4, the additive relationship of paternal halfsibs.
The subsequent approximate standard errors are expected to underestimate the

actual standard errors (Swiger et al., 1964).

4. RESULTS AND DISCUSSION

4.1 Model 1 Genetic Parameters

The estimates of sire variance, environmental variance, and heritability for birth
weight (BW), adjusted weaning weight (WW), adjusted yearling weight (YW), post-
weaning average daily gain (ADG), end of test weight (EOTWT), carcass weight
(CARCWT), marbling score (MARB), longissimus muscle area (LMA), 12th rib fat
thickness (FAT), carcass grade (CARCGR), and cutability (CUT) from Model 1 are
reported in Table 11. This procedure is not capable of modeling continuous variables, so
covariates were not included in the model. Additionally, no sire relationship information
was included in the Model 1 analysis. With the exception of adjusted yearling weight
heritability, all other traits are below the mean of heritability values listed in Table 1. It
was expected that the estimates of sire variance would be initially low ﬁom this single trait
model due to the limitation of the eﬁ‘ects modeled, in addition to the biases that occur
ﬁom selection on correlated traits to the single trait in the model. However, initial

variance components were obtained.

4.2 Model 2 Genetic Parameters

The results of Model 2 are included in Table 12. Estimates achieved ﬁom Model 2

are ﬁ'om single trait mixed models, and therefore could be subject to bias from sequential

62

63

 

 

44.4 444.2 44 2 .4 42222224226
44.4 444.4 424.4 422420 448240
44.4 444.4 444.4 444524222. 24.2 .2222 22242
44.4 442 . 22. 422.4 442 4244422 44422442222242
2 2.4 424.4 224.4 4284 4222222242
42 .4 2.44.444 444.44 2222224? 448.240
44.4 2.44.44: 444.442 2.2424? 244.4 .24 222242
44.4 444.4 444.4 22240 422422 444242 42224834442
24.4 444.4442 444.442 2224243 44222824 224244.222
2 24 444.444 444.: 22222243 22222483 224244.22
44.4 444.44 444.4 24243 2222222
42222224222442 4444224> 24222422222422.2442 4422422; 4224 22424

 

SEE-32.22— .2224 432242.22; 32228.2 432242.22; 0.224 ..e 4848228 4 .232 .3 034,—.

64

 

 

44.4 424.2 444.2 4424 4222224240
44.4 444.4 444.4 424.4 422420 448.240
44.4 444.4 444.4 444.4 48524424 242 4222 22242
44.4 444.44 444.42. 444.4 82 4244222 42222224442442
44.4 444.4 444.4 444.4 42444 224224242
42 .4 424.444 444.444 444.: 2224243 4.4.8240
42 .4 2 24.2442 444.2 242 444.44 224243 28424 222242
24.4 444.4 444.4 444.4 4240 422422 444242 442483-284
42 .4 444.4442 444.4442 444.44 22222243 44222824 224244.242
2 24 444.444 444244 444.42 224243 4222483 442422.22
44.4 444.44 444.44 424.2 22222243 2222222
4222244222422 4444.44.» 42224242242242 44442242, 24222422222422.2422 4422422424 4.224 22424

 

642242.242, 42942222..— .2224 4322—2249. 44.224 .2224 52:28:92. .3 434222228 u .232 .m— 034,—.

65

selection and also bias ﬁ'om selection on correlated traits (Pollak et al., 1984; Walter and
Mac, 1983). The heritability values reported in Table 12 are lower than the mean
literature cited estimates listed in Table 1. This model attempted to remove the biases
from omitting genetic relationship information through the additive relationship matrix, in
addition to the effects of slaughtering the cattle at diﬁ‘erent ages. A comparison of Model
1 and Model 2 indicate that there was an increase in the sire variance and heritabilities of
carcass traits estimated ﬁom Model 2. Although there was a decrease in the heritability
estimates for the growth traits, a review of Henderson (197 5) suggests that Model 1 could
be biased due to the omission of relevant factors of the model. Additionally, Model 2
showed greater differences in the heritability values for both adjusted yearling weight and
marbling score. Adjusted yearling weight heritability ﬁ'om Model 2 could be lower due to
modeling the covariance associated with the sire additive genetic relationship matrix.
Marbling score might be higher in Model 2 for this same reason, in addition to eﬂ‘ects
associated with possible removal of environmental variation from slaughtering cattle at

different ages.

4.3 Model 3 Genetic Parameters

Table 13, Table 14, and Table 15 contain the genetic, environmental, and
phenotypic variance estimates, respectively, from each ﬁve trait combination for Model 3.
Table 16 contains the heritability estimates ﬁom each ﬁve trait combination of Model 3.
Tables 17, 18, and 19 contain the mean, standard deviation, and range of the genetic,

environmental, and phenotypic variances obtained from each ﬁve trait analysis of Model 3.

66

424224220 2422322424 24 ow4 224 22223 220444322444 4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

442.4 442.4 442 .4 442.4 . 22222224226
424.4 2.244 424.4 2.24.4 . 42280 248240
444.4 444.4 444.4 444.4 4 2448242224 2442 22222 22242
C
444.4 2.44.4 444.4 244.4 82 42442242 44222244242242
444.4 444.4 444.4 244.4 . 4284 4222222242

444.44 444.44 424.44 444.44 22222243248240
444.44 444.44 444.44 442 .42. 2224243 282. 24 22442

22240422442
444.4 444.4 444.4 444.4 4224242 42222282424442

444.44 244.44 444.44 444.44 2ﬁ2ww.wﬁ2w&wﬁmma«z

. .... 44,442 444.42 424.42 244.42 444.42 424.42 424.42 444.42 444.42 444.42 22242434424832.4244 .22...-

424.2 424.2 444.2 444.2
42 4 4 4 4 4 4 4 4 2

42442422 42442422 42442422 42442422 424.2242 4.2442422 224.2242 424.2242 4.2442422 22242422 22422.

 

 

 

.343224 22222242232228 224.22 22>..— .3 43 2242...; 422224 n .2222: .3 034,—.

 

 

242.298 4822922424 44 emu 224 55 40444292 .

 

67

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

442.2 444.2 444.2 444.2 . 44234.5
444.4 444.4 444.4 444.4 . 4445 4485
444.4 444.4 444.4 444.4 . 48522224 .44 .2242 2442
I
44.24 444.44 444.44 444.24 82 428422 444224444223
444.4 444.4 444.4 444.4 . 284 44224442
442.424 444.444 442.444 444.424 ... 4.22.4443 448:6
444.4442 444.4442 444.4442 444.4442 42242443444444.4442...
44.444.424.42-
444.4 444.4 444.4 444.4 448434 442483.444
422.4442 444.4442 444.4442 444.4442 .mmwwﬁwmmmmmwmmwmﬁﬁ,
444.244 444.244 444.244 444.444 244.244 444.244 444.444 424444 444.244 444.244 ..ﬁ§4§§$§€¢-
444.44 444.44 44.4.44 444.44 2224532244242
42 4 4 4 4 4 4 4 4 2
44442454 4442422 44442422 44442422 4.2442422 42442422 44442422 4442422 44442454 42442422 .284

 

 

.444422224 223442234223 :22... 9:.— ..o 842242.223 Eunoﬁggiao n .252 .3 034,—.

 

69

0442498 482322424 24 own :4 22:49 34442454 .

 

Nm.0

mmd

N00

and

. 44222435

 

NN.0

NN.0

mad

mN.0

. 2.440 44830

 

44.0

30

ﬁnd

5.0

. 48524424 4.4..2 4:2 .442

 

wmd

0N0

and

and

I
82 028222 38_8&:3

 

0N0

5N0

5N0

0N0

. 284 4222242242

 

2.0

3.0

2.0

:0

. 424243 44.85

 

3.0

0:0

3.0

2.0

424243 484 44 422m

 

:N.0

:N.0

0N0

vmd

222240 NEE

omﬂo>< $2083.63

 

3.0

0:0

0:0

:0

4224243 422224.344 242424.434

 

:.0

:.0

:.0

0:0

:.0

:.0

0:0

0:0

:.0

00.0

3303 Macao? 024.22.8«4

 

NN.0

NN.0

NN.0

NN.0

4224443 22542

 

 

42
4244242244

 

44442422

 

4
42442422

 

44442422

 

m

 

4.4442454

 

.8032: 222442232228 4.24.24 9»: 44 482222:

 

 

4
444.4%

 

m

FE

 

 

4
42442422

 

 

2
424324224

who

 

4:44P

NEE-43.22. m $.52 .3 «Ban.

 

70

624.550 482322424 24 own :4 2E? coax—.4224 .

 

 

444.4 442.4 442.4 44444.4 442.4 . b24435
4244.4 2424.4 4424.4 44444.4 424.4 . 4480 444240
884 4444.4 4444.4 44444.4 444.4 . 444524424 4442 4242 22.42
444.4 244.4 444.4 442 .4 444.4 . 82 42842 4484444443
4444.4 4244.4 4444.4 42 244.4 444.4 . 284 4422424:
244.4 444.44 444.44 444.4 442.44 . 4.24443 4.44240
444.42 444.44 442.44 244.4 444.44 42443 4444 44 4442
4444.4 4444.4 4444.4 42444.4 444.4 55 424442 448434 452.443-4444
444.42 444.44 444.44 442 .4 442 .44 444443 442244444 4844.234
444.4 444.42 444.42 442.2 444.42 424443 45483 4244.23
444.4 444.2 444.2 4444.4 444.2 24243 5442
44252 545432 8444242 82.44642 2:42.434 :82 444.4

 

 

34.2538 323...; 840.2% 44 owaﬂ E224 £343.50 Eat—284 .2232: n .232 .4.— 0.4—22h.

71

34244460 48534—4 44 ow4 :4 223 35424224 .

 

 

424.4 444.2 444.2 4444.4 444.2 . 444435
444.4 444.4 444.4 4444.4 444.4 . 4445 448440
4444.4 4444.4 4444.4 2444.4 444.4 . 444524424 E2 442 542
44.2 444.44 444. 24 444.4 444.44 . 32 42842 4242444443
2444 444.4 444.4 4444.4 444.4 . 284 4444.422
42 2.44 444.444 442.424 424.44 424.244 . 42443 444240
444.242 444.4442 444.4442 444.44 4444242 4224443 4444 44 4442
4444.4 4444.4 4444.4 4444.4 444.4 .2240 44442 448434 454443.444
444.442 444.4442 444.4442 444.44 444.4442 424443 4422.44.» 4244.234
444.44 444.244 444.244 244.4 444.444 424243 444433 4844.434
444.4 444.44 444.44 444.4 444.44 4.24443 5444
44.52 8454a: 8244422 €444,442 254454 4442 444.4

 

5448448 32242.24.» .44—22.22.92.224 .44 «M224.— ..224 .424-430.. 242.2234 .223... n .2442 .2 9.4.4:.

72

34244448 48232224 44 ow4 224 233 34424224 .

 

 

44.4 244.2 244.2 4444.4 444.2 . 4444446
2444 444.4 444.4 4444.4 444.4 . 4445 444840
4444.4 4444.4 4244.4 4444.4 444.4 . 44452424 4442 4:2 2442
444.2 444.44 444.44 444.4 444.44 . 82 424442 4444444443
2444 444.4 444.4 4444.4 444.4 . 284 422224.42
424. 24 444.444 424.444 444.44 444.444 . 4224443 44426
444.442 444.4442 424.4442 444.44 424.4442 424443 4444 44 44m
4444.4 4444.4 4444.4 2444.4 444.4 540 44442 444434. 442483-484
444.242 444.4442 444.4442 244.44 442.4242 444443 4444484 4344.234
444.42 444.444 444.444 444.4 44 2 .444 444443 442483 4344.234
442 .4 444.44 444.44 4444.4 442.44 42443 4:242
44.232 8424482 5244492 44444542 4.444444 44422 444.4

 

5445.48 3224...: 93.444.22.42 .44 ans... .4224 £343.50 ...—4.42.34 .2240:— n .332 .3 934,—.

73

342.2454 482342424 44 $4 :4 22443 00444444244 4

4842424440 : 24422 2.423 22303 $242240? 3444284 244848 44442 .005 40.4 48484444 442—5422.224 428.4 0.203
82E. 22.6 4244422444 84824840424 4.8425244 442—3444.24: 4222 .40 42408 2: 4224 842424444 42:54:40: 05 .40 2240222 0222 4542228 228200 .

 

 

424.4 444.4 444.4 4444.4 244.4 4 44.4 4 4444446
4244 444.4 424.4 4444.4 244.4 4 44.4 4 4445 44426
444.4 444.4 444.4 4244.4 444.4 4 44.4 4 444524444 .2 .44 .442
424.4 444.4 444.4 4444.4 244.4 4 44.4 ._ 442 424442 4244444443
4244 444.4 444.4 4444.4 244.4 4 44.4 ._ 284 44442422
4244 442 .4 442 .4 4444.4 444.4 4 42 .4 ._ 444443 44428
444.4 442 .4 442.4 4424.4 244.4 4 42 .4 424443 444 44 44m
444.4 444.4 444.4 4424.4 244.4 4 24.4 .440 44442 444.434 44483.4444
424.4 442 .4 442 .4 4444.4 244.4 4 42 .4 424443 444.484 4844.244.
444.4 42 2 .4 444.4 4444.4 444.4 4 2 2.4 424443 4424443 4244.234
444.4 444.4 424.4 4444.4 444.4 4 44.4 424443 224442
44432 84422442 424444424 344446 44:4 . 484,2 42.4
44443442 4444434

 

6644242448 34.44.444.22. .44 own”. 422:. 542444.524 2:422:34 .2242: n .0232 .3 24.4.4.

74

Table 20 contains the mean of the heritability estimate and the mean of the heritability
estimate’s approximate standard error, in addition to the standard deviation of the
heritability and range for each trait. Inspection of the standard deviations and ranges in
Tables 17, 18, and 19, respectively, indicate the largest deviation of the variance estimates
occurred for adjusted yearling weight and end of test weight. A reason for this is there is
more genetic variation for this trait within the Canadian Charolais population. Also,
greater environmental variation was estimated to exist in these two traits, possibly due to
environmental inﬂuences which had more time to oppose existing genetic factors.
Although there was deviation in the variance estimates, no range of the heritability
estimates exceeded the approximate standard errors, which are expected to be

underestimated (Swiger et al., 1964).

4.3.1 Heritability Estimates

Average heritability estimates for the traits analyzed are presented in Table 20.
The heritability for birth weight ﬁ'om this study was 0.22 i: 0.078. Other studies have
reported birth weight heritability to be moderate to highly heritable. This value is lower
than the average of 172 estimates by Koots et al. (1994a) of 0.35 for birth weight
heritability. However, it is in close agreement with the value reported by Johnston et al.
(1992) of 0.25 that was attained from the Canadian Charolais Association’s Conception to
Consumer program data, and is higher than the estimate of 0.18 found by Veseth et al. et

al. (1993) on 736 Hereford bulls.

75

The average heritability estimate for W was 0.11 i 0.079. Previous estimates
cited from the literature indicate that W is a low to moderately heritable trait. Koots et
al. (1994a) reported an average heritability value of 0.27 from 239 literature sources. The
present study’s heritability value is in agreement with other investigations that have shown
WW heritability values of 0.09, 0.09, and 0.18, respectively (Arnold et al., 1991; Johnston
et al., 1992; Woodward et al., 1992), but is in the low range reported in Table 1.

The heritability value discovered in this analysis for YW was 0.19 i 0.081. Koots
et al. (1994a) after reviewing 154 papers detected YW heritability to be moderately
heritable with a mean value of 0.35. This study’s results, however, are in agreement with
other published results of 0. l4, and 0.16, respectively (Arnold et al., 1991; Johnston et al.,
1992) and also the 0.25 estimate reported by de Rose (1992) from Charolais data obtained
from the Canadian Beef Sire Evaluation Program.

Koots et al. (1994a) analyzed 24 heritability estimates for ADG and reported a
mean value of 0.24. Other heritability estimates in the literature indicate that ADG is
moderate to highly heritable, with values of 0.13, 0.19, 0.47, and 0.52 (Arnold et al.,
1991; Johnston et al., 1992; Veseth et al., 1993; Benyshek, 1981). The results of this
analysis indicate a heritability value for ADG of 0.21 i 0.081, which is within the range of
estimates throughout the literature.

The heritability estimate for EOTWT was 0.18 i 0.081. Other cited heritability
estimates for EOTWT in the literature include 0.42, and 0.52 ﬁ'om Veseth et al. (1993)
and Benyshek (1981), respectively. Koots et al. (1994a) analyzed twelve EOTWT

heritability estimates that had a mean value of 0.37. The EO'I'W'I’ heritability estimates in

76

the literature suggest that EOTWT is moderate to highly heritable. Therefore, this study
yielded a heritability value that was low for this trait, which could be due to reduced
variation in the reduced population of sires that possibly were selected for grth traits.

Koots et al. (1994a) reported a mean heritability value of 0.45 from 19 literature
sources for hot carcass weight that were estimated at an age constant slaughter basis.
Additional references for age constant carcass weight include 0.31, 0.38, 0.41, and 0.43
from Mlson et al. (1993), Veseth et al. (1993), Marshall (1994), and Koch et a1. (1982),
respectively. This study’s heritability of 0.13 i 0.082, which is below the moderate and
high heritability values indicated throughout the literature. The apparent low heritability
values for the various growth traits, including age constant carcass weight, may be due to
the selected sample of sires used in the Conception to Consumer program, which had a
tendency to be grth bulls. Robertson (1977) reported that if selection on the parents is
based on the trait on which heritability is being measured then the estimates may be biased
downward due to reduced additive genetic variance of the parents.

Heritability values of 0.27 i 0.081 for MARB were estimated from these data.
Koots et al. (1994a) estimated heritability values of 0.37 for age constant marbling score,
with the estimate being the mean of 12 estimates. Other heritability values cited
throughout the literature for MARB include 0.23, 0.26, 0.31, 0.31, and 0.35 (Woodward
et al., 1992; Wilson et al., 1993; Cundiffet al., 1971; Veseth et al., 1993; Marshall, 1994),
respectively. The referenced heritability estimates for MARB indicate that the trait is

moderate to highly heritable, and these data show a similar conclusion.

77

This study estimated a heritability value of 0.29 i 0.081 for LMA. Other
heritability values for LMA in the literature include 0.32, 0.37, 0.40, and 0.41 from Wilson
et al. (1993), Marshall (1994), Benyshek (1981), and Cundiﬂ‘ et al. (1971), respectively.
Koots et al. (1994a) found 16 LMA estimates to average 0.43. The literature values
suggest that LMA is a moderate to highly heritable trait, which would indicate that the
present study’s LMA estimate is at the low range of cited estimates.

Heritability values reported in the literature for FAT include 0.26, 0.41, 0.44, 0.49,
and 0.52 (Veseth et al., 1993; Koch et al., 1982; Marshall, 1994; Arnold et al., 1991; and
Benyshek, 1981), respectively. Koots et al. (1994a) reported 26 literature references for
FAT that had a mean heritability value of 0.43. The current study’s FAT heritability value
of 0.37 :t 0.080 indicates that the FAT heritability value ﬁts into the low range of
reported heritability values. The age constant 12th rib fat thickness indicates differences in
the rate of fat deposition, which is affected by the age that an animal reaches physiological
maturity. Carcass weight constant 12th rib fat thickness predicts differences in the volume
of fat thickness, thereby indicating at what size physiological maturity is reached.

The present study yielded a heritability estimate for CARCGR of 0.23 i 0.081.
There were no references detected in the literature for carcass grade under the Canadian
grading system, in part because the carcass grading standards were revamped in 1987, and
also because many of the major component traits for carcass grade have been previously
analyzed.

Koots et al. ( 1994a) reported 12 heritability estimates for CUT to have a mean

value of 0.41. Additional literature estimates include 0.18, 0.28, 0.36, and 0.49 from

78

Woodward et al. (1992), Cundiff et al. (1971), Marshall (1994) and Benyshek (1981),
respectively. These literature estimates indicate a considerable range for heritability
estimates, although they indicate for the most part that CUT is moderate to highly
heritable. The present research resulted in a heritability value of 0.32 i 0.081 for CUT.
This result appears to be well within the range of reported values for CUT.

The deviations associated with the variance estimates, in addition to relatively low
heritability estimates compared to literature values, certainly have positive implications.
The Charolais bulls used in this study were a group selected for high growth, which the
current data indicate variability exists within these growth traits. In a breeding program,
Charolais breeders certainly have more opportunity to select breeding stock at the
extremes of their given traits of interest which will allow for more rapid improvement. If
no variation existed within the population, no progress can be made, as breeding stock
would produce offspring that exhibit the same performance as the parents.

These data suggest that the grth and performance traits of birth weight,
adjusted weaning weight, adjusted yearling weight, post-weaning average daily gain, end
of test weight, and even carcass weight, are low to moderately heritable. Carcass traits
analyzed including marbling score, longissimus muscle area, 12th rib fat thickness, carcass
grade, and cutability were moderately heritable. Higher heritability values allow breeders
to make more accurate decisions when selection occurs using estimates of genetic merit of
livestock. Lower heritability values indicate more progress might be made for a trait

through changes in the animal’ s environment.

79

 

 

 

Table 21.
Model 3 averametic and environmental correlations for growth traits " ”.
BW ww YW ADG EOTWT
Buth Weight 022‘. 0.40 I 0.32 I 0.18 I 0.24
I I I l
___________________________ I ____I _____I ____I _______4
Adjusted WeaningWeight 034 T011 T073 T034 T 076 J
___________________________ I ____I ____I ____I _______
Adjusted Yearling Weight 022 T 067 T019 T 083 T 083
___________________________ I _____I ____I ____| _______
Post-weaning Average Daily Gain 0.14 T 0.19 T061 T021 T 0.94
_____________________ _‘_____| ____I ____I ____I _______
End OfTestWeight 0.34 T064 T058 T039 T 018
I I I l

 

 

 

 

‘ Average genetic correlations above the diagonal, average environmental correlations
below the diagonal, heritabilities on the diagonal.

b BW = Birth Weight; W = Adjusted Weaning Weight; YW = Adjusted Yearling
Weight; ADG = Post-weaning Average Daily Gain; EOTWT = End of Test Weight.

80

Table 22.

Model 3 average_genetic and environmental correlations for carcass traits " b“.

 

 

 

 

 

CARCWT MARB LMA FAT CARCGR CUT
CarcassWeight 0.13 I -019 I 018 I 017I 0.17 I 0.01
------------------- T-'---T-'--T"'""T""""""T“ ""
Mar-11.1111.S_S.<29£e._---_-_----9-9§__-1--9-21-4.- 0- 1.6 ...-1- -0_ 32 4..-:992--,1_9-3-4-.
”Lansiﬁupyali’lgeslzétea----Q-.52-_,1.-9-11-.;.-0.2_9 _-,1_ -0- 24 -.1___:9-12--1 9.8.5..
-1-2-t11131-11_F_at-I111£191£§_S---___9-9§-_1--0-1.6- 1 -0_13 1 0.3-7 1 _-Q-§7_-_1 ;0_-6_7..
34242901312 ....... .__-_9-..0:1-_T--.<.1-0-7-T- -0 1.7. T .0: (1:3 _T-_1.1.-Z3.--1 51-90..
Cutability 0.34 T 0.16 . 0.82 T0 62 T -0.45 T 0.32

 

 

‘ Average genetic correlations above the diagonal, average environmental correlations

below the diagonal, heritabilities on the diagonal.

" CARCWT = Carcass Weight; MARB = Marbling Score; LMA = Longissimus Muscle
Area; FAT = 12th Rib Fat Thickness; CARCGR = Carcass Grade; CUT = Cutability.
° Carcass traits were analyzed with an age at slaughter covariate.

624424.60 2822022424 24 ow4 224 2222.: 2204424224 9295 422422 4440.240 .

522342220 u 50 622420

42.28240 .1. 220042.440 24422220225. 2442 252 2222 n .5222 240244 0282222 42222224420223 u <22 M42.243 02223242 n 522 2222024? 4444.240 .1. 2.3920 222202045
284 24 222242 1 2.3.4022 222422 422442 4422242 44222834422103 2224243 422222824 442422.431; 2224243 42222283 224244.232" >43 4224243 22222421342 .-
.24220m4222 0222 224 42222222222422.2022 24224w42w 4222 323 422022424228 24222022222022.4224 ow42o>4 2425m4€ 4222 9604 422022424228 22224» ow424>< .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

44.4 n _ n n r .4 444 _ 42.4 m 44.4 . 44.4 _ 44.4 24222224240

H 44.4 2. n P .2 .4 244. H 24.4 4 24.4.4 444. #42. 4. 4244401414414w4w1

" L. 44.44. H n P. 44.4 H444 “ 44.4 ﬁ 24.4.i424. 4m4142124422w412®1421~21424121

” u .. 44.4 .- _4 H 44.4 H 44.4 2 44.4P444 " 44.4 441%42444221414412221241424244021

H .. L. P 44.4 2 H 44.4 n 44.4 " 44.4H4444 44.4 4128mm2m2m2w2m4m

_ 2. 2. H n 42.4 H 44.4 n 44.4 2 444 2 444 _ 444 22244.43 448240

44.4. n 44.4 _ 44.4 H 22.4 n 44.4. 2 44.4 m 42.4 _ u 2 _ 22wmm212memwnwm4w1

2 _ _ _ _ _ _ . _ - 222I4Wl

44.4. n 42.4 n 42.4 n 44.4 n 44.4. n 44.4 I. n 24.4 n n n 422424442432 4222224432444

44.4.2- 444 n 44.4 n 42.4 n 42.4. n 44.4 I. n _ 42.4 n “ 2.242434222224424 2242422421224

44.4. _. 24.4 n 24.4.. 44.4 F 22.24 _ 44.4 n P 2 L- 224 .1 2224243 42224443 4m214m4wm41

24.4 _ 44.4. _ 44.4.” 44.4 2 42.4 _ 44.4 n _ 2 _ 2 44.4 22242444 2222242
2.00 228.225 4<..2 <22 42422 56,225 2.32.042 oo< >25 33 342

...2 .. 42222.22 44424.23 42224 2.234.234.5242. 442422429228 .422222222422224 42mm 422224 a 4.2422 m .3242 .m~ oEalHW

 

82

4.3.2 Genetic Correlations

With highly positive genetic correlations between any given traits X and Y, bulls
that excel in trait X are also expected to produce progeny that excel in trait Y. Likewise,
if the genetic correlation is highly negative between traits X and Y, bulls that excel in trait
X are expected to sire oﬂ‘spring that are inferior for trait Y, respectively. Therefore,
genetic correlations provide an indication of response due to selection to producers when
heritability values are similar and selection intensity is held constant. The average genetic
correlations among growth traits, among carcass traits, and among growth and carcass
traits are included in Tables 21, 22, and 23, respectively.

This study’s genetic correlations between BW and WW and BW and YW were
0.40 and 0.32, respectively. Woodward et al. (1992) analyzed data from the American
Simmental Association, in addition to other studies in the literature from Veseth et al.
(1993); and Koots et al. (1994b) reported genetic correlations between BW and W of
0.33, 0.54, and 0.47, respectively. Koots et al. (1994b) provided a similar mean genetic
correlation from 37 sources between BW and YW of 0.48. The current study is in
agreement with the published results, with only a minor decrease in magnitude for genetic
correlation between these trait combinations.

Koots et al. (1994b) reported 66 genetic correlation estimates between WW and
YW, in addition to W and ADG to have a mean value of 0.78, and 0.39, respectively.
This study detected a genetic correlation between WW and YW of 0.73, along with a
genetic correlation between WW and ADG of 0.34, respectively. This study agrees with

results published in that selection for increased WW will result in a highly correlated

83

response in YW, while still achieving a moderate increase in ADG. The genetic
correlations between the post-weaning growth traits of YW, ADG, and EOTWT are all
high ( 2 0.83) and in agreement with the magnitudes from results published by Koch et al.
(1982), Veseth et al. (1993), and Koots et al. ( 1994b).

A genetic correlation between MARB and LMA of 0.16 was discovered from
these data. Selection for improved marbling should yield lighter muscled cattle. These
results, which have the highest marbling score numerically the lowest, are in agreement
with the respective reported values of -0.12, -0.14, -0.14, and -0.23 (Brackelsberg et al.,
1971; Koch et al., 1982; Marshall, 1994; Koots et al., 1994b), but yet do diﬁ‘er ﬁ'om the
results of Van Vleck et al. (1992) and also Veseth et al. (1993) who reported genetic
correlations of -0.40 and 0. 51 between MARB and LMA.

This study additionally found the genetic correlations between MARB in
combination with FAT and also CUT respectively of -0.32 and 0.34. Marshall (1994) and
Koots et al. (1994b) report similar genetic correlations between MARB and FAT of 0.37
and 0.36, respectively. Additionally, these same published results indicated a genetic
correlation between MARB and CUT of -0.24 and -0.54, respectively. This suggests that
selection for improved marbling scores will result in a moderately correlated increase in
FAT and a decrease in CUT.

A genetic correlation of -0.60 was found between CARCGR and CUT. As is
expected in the Canadian meat grading system, an improvement in carcass grade will result
in a highly correlated increase in cutability. Furthermore, a genetic correlation between

LMA and FAT along with LMA and CARCGR was -0.24 and -0.19, respectively. The

84

genetic correlation between LMA and FAT is intermediate to the published results of
-0.06, -0.08, -0.09, -0.37, and -O.44, respectively (Wilson et al., 1993; Koots et al., 1994b;
Brackelsberg et al., 1971; Arnold et al., 1991; Koch et al., 1982). These results indicate
that selection for heavier muscled cattle should result in a low to moderate improvement in
fat thickness and also carcass grade. Furthermore, this study’s genetic correlation
between LMA and CUT is 0.85. Marshall (1994) reported a correlation of 0.53, and
Koots et al. (1994b) found three papers that had a mean genetic correlation for LMA and
CUT of 0.26. Moreover, the genetic correlation between FAT and CUT was -0.67. This
high correlation is intermediate to the genetic correlations reported by Cundiﬁ‘ et al.
(1971) and Koots et al. (1994b) of -0.74 and -0.33, respectively. In addition, the genetic
correlation between FAT and CARCGR was discovered to be 0.87. These high genetic
correlations between CUT and LMA along with FAT, in addition to CARCGR and FAT
can be expected, as they are component traits of CARCGR.

Genetic correlations between BW and CARCWT, and BW and LMA of 0.27 and
0.22, respectively. When analyzing the genetic correlation between BW and CARCWT,
Koch et al. (1982) and Veseth et al. (1993) detail correlations of 0.60 and 0.11,
respectively. The same authors also reported genetic correlations between BW and LMA
of 0.31 and 0.57, respectively. The results of this paper are in agreement with these cited
estimates that the genetic correlations between BW and CARCWT, and BW and LMA are
positive and moderately correlated, although this study’s results are slightly lower in
magnitude. A high correlation of 0.76 was detected between WW and CARCWT. Other

genetic correlations found in the literature between WW and CARCWT include 0.13,

85

0.82, and 0.84, respectively (Arnold et al., 1991; Marshall, 1994; Koots et al., 1994b).
This study agrees with results published in such that selection for increased WW will result
in a highly correlated response in CARCWT.

The genetic correlation between BW and MARB was found to be 0.17, while a
genetic correlation of -0.04 was found between BW and FAT. Koch et al. (1982)
reported a genetic correlation between BW and marbling score of 0.31, while Woodward
et al. (1992) and Veseth et al. (1993) reported correlations of the same trait combination
of 0.05, and -0.18, respectively. As the marbling score scale is reversed in the present
study, these results are extremely close to the correlation found by Veseth et al. (1993),
such that selection for lower birth weights will result in increased marbling. The genetic
correlation between BW and FAT found in this study of -0.04, although it has the same
sign as the -0.27 correlation found by Koch et al. (1982), is lower in magnitude, and
would indicate that selection for reduced birth weights will yield only a low correlated
increase in FAT.

The genetic correlations between the post-weaning growth traits of YW, ADG,
EOTWT, and also CARCWT are all high ( 2 0.83) and in agreement with the magnitudes
from results published by Koch et al. (1982), Veseth et al. (1993), and Koots et al.
(1994b). These traits (YW, ADG, EOTWT, CARCWT) also have very similar genetic
correlations with the carcass traits MARB (range = -0.08 to -0.26), LMA (range = 0.09 to

0.18), FAT (range = 0.15 to 0.28), CARCGR (range = 0.15 to 0.24), and CUT (range

— -0.09 to 0.01). Koch et al. (1992) was in agreement with these ranges where

applicable, but reported higher correlations for the combinations with LMA (range = 0.34

86

to 0.44). Arnold et al. (1991) had mutual results with the exception of ADG x MARB
(0.54), CARCWT x MARB (0.33), and CARCWT x FAT (0.36). Additionally, Arnold et
al. ( 1991) also found negative genetic correlations for YW and ADG in combination with
LMA, -0.06, -0.18, respectively. The results of Marshall (1994) and also Koots et al.
(1994b) support these ranges in most instances, with the exception of the combinations

with LMA and CUT, which had more extreme genetic correlations.

4.3.3 Environmental Correlations

Environmental correlation estimates are presented with the genetic correlation
estimates between growth traits analyzed in Model 3 in Table 20. Birth weight yielded a
low environmental correlation to post-weaning average daily gain (0.14), but was
moderately correlated to the other growth traits with a correlation of 0.34 to adjusted
weaning weight and also end of test weight. Adjusted weaning weight possessed a low
environmental correlation to post-weaning average daily gain of 0.19. However, adjusted
weaning weight, adjusted yearling weight, and also end of test weight all had high
environmental correlations of 0.67, 0.64, and 0.58. This suggests that environmental
inﬂuences, which include feedlot conditions and diet, tends to impact these traits so that
cattle that had high adjusted weaning weights tended to also have high adjusted yearling
weights and end of test weights.

Environmental correlations between carcass traits are included in Table 22.

Carcass weight had a high environmental correlation to longissimus muscle area, but a low

87

environmental correlation to other carcass traits. The -0.04 environmental correlation
between carcass weight and carcass grade is desirable, which due to the carcass grade
scale indicates that cattle that had heavier carcass weights also possessed advantageous
carcass grades. Marbling score had only low environmental correlations to other carcass
traits; but yet as marbling score goes down numerically, environmental inﬂuence yields a
correlated increase in 12th rib fat thickness (-0.16 environmental correlation). There was
a high environmental correlation between longissimus muscle area and cutability (0.82).
Moreover, carcass grade had a medium environmental correlation to cutability as well as a
high positive environmental correlation to 12th rib fat thickness.

Environmental correlations between the grth and carcass traits are included in
Table 24. Carcass grade, 12th rib fat thickness, and marbling score all had low
environmental correlations to every growth trait modeled. Cutability and also longissimus
muscle area were moderately environmentally correlated to growth traits, but carcass

weight had high environmental correlations to growth traits.

4.3.4 Phenotypic Correlations

Phenotypic correlations are the result of the combination of eﬁ‘ects from
environmental variation and also the genetic variation. The phenotypic correlation
formula can be shown as (Scarle, 1951):

r, = 1,0..112)‘ + arm-1100412)]-5

88

where:

rp is the phenotypic correlation,

r, is the genetic correlation,

r. is the environmental correlation,

h. is the heritability of trait one, and

h; is the heritability of trait two.

If the ratio of the environmental correlation to the genetic correlation exceeds the ratio
[1-(h1h2)"]/ [(1-h.)(1-h2)]", then the phenotypic correlation exceeds the genetic
correlation. When this occurs, the phenotypic correlation will then follow the sign of the
environmental correlation. This discussion serves to show the magnitude that lowly
heritable traits have on inﬂuencing phenotypic correlations. If the discussed traits one and
two possess high heritabilities, then the effect of environmental correlation is reduced, as
the denominator of this ratio is reduced. But when trait one and trait two heritabilities are
low, the denominator of the ratio is increased, and the environmental correlation is
multiplied by a larger value in the phenotypic correlation equation.

Phenotypic correlations between the grth traits analyzed in Model 3 are
presented in Table 24. Birth weight had only low to moderate phenotypic correlations to
other growth traits. Adjusted weaning weight, adjusted yearling weight, and also end of
test weight had high phenotypic correlations amongst each trait. This suggests that cattle
that have high adjusted weaning weights also tended to have high adjusted yearling and

end of test weights. These results are in general agreement with the estimates cited in the

89

literature. Post-weaning average daily gain had a high phenotypic correlation of 0.67 to
adjusted yearling weight, but had a low phenotypic correlation to both birth weight and
also adjusted weaning weight.

Phenotypic correlations between carcass traits are included in Table 25. Carcass
weight had minimal phenotypic correlations to other carcass traits, with the exception of
cutability, and also longissimus muscle area, which are moderate and highly phenotypically
correlated, respectively. Cundiﬂ‘ et al. (1971) reported a phenotypic correlation between
carcass weight and cutability of -0.44 on British breed steers, while Koots et al. (1994b)
found three phenotypic correlations to have a mean of -0.03. These literature values differ
from the 0.34 phenotypic correlation found in this study.

Carcass weight expressed a genetic correlation to marbling score of -0. 19, but
only a correlation of 0.05 and 0.04 for environmental and phenotypic correlations,
respectively. This would indicate that selection for carcass weight using breeding values
would result in more marbling, but the environmental effects that cause carcass weight to
increase also cause less marbling. Marbling score possessed a moderately negative genetic
correlation to 12th rib fat thickness (-0.32), yet only expressed a phenotypic correlation of
-0.17. Similar results occurred between marbling score and cutability, indicating more
desirable results should be achieved through selection using breeding values. This study
analyzed a phenotypic correlation between cutability and longissimus muscle area in
addition to cutability and 12th rib fat thickness of 0.82 and -0.62, respectively. These
values are higher than the cited estimates of 0.45 and -0.36 reported by Marshall (1994),

and Cundiff et al. (1971), respectively.

90

Phenotypic correlations between the grth and carcass traits are included in
Table 26. Carcass weight had high phenotypic correlations to all growth traits, in addition
to the results of this study being greater in magnitude than the average correlations of
carcass weight to grth traits listed in Table 7. Nevertheless, marbling score, 12th rib fat
thickness, and carcass grade all had low phenotypic correlations to growth traits. If
producers use breeding values to select for improved marbling, they should achieve
desirable results in correlated grth traits with the exception of birth weight. However,
the importance of using breeding values in selection is apparent as the phenotypic
correlations between marbling score and grth traits causes antagonistic responses.
These same results were concluded by Koots et al. (1994b) in their mass review of
published estimates. Longissimus muscle area possessed moderate correlations to all
growth traits, as was expected by the literature phenotypic correlations.

When producers place no genetic selection either on carcass grade or cutability,
they will achieve desirable responses in correlated grth traits due to favorable
environmental inﬂuences. However, if breeders select for either carcass grade or cutability
on genetic merit, they should observe antagonistic results in correlated growth traits.
These correlations of antagonistic traits are important for breeders to note, so they can
identify genetic sources that do not follow these results, therefore making simultaneous

improvement in both traits.

91

4.4 Model 4 Genetic Parameters

A fourth analysis was run to look at the variation in marbling score, longissimus
muscle area, and 12th rib fat thickness whole holding carcass weight constant to observe
differences in marbling score, longissimus muscle area, and 12th rib fat thickness as

proportions of the carcass (Cundiff et al., 1971).

4.4.1 Heritability Estimates

Model 4 heritability estimates are presented in Table 27, in addition to the Model 4
variance components. Koots et al. (1994a) estimated heritability values of 0.37 for
carcass weight constant marbling score, with the estimate being the mean of four
estimates. Heritability values of 0.28 :t 0.081 for carcass weight constant marbling score
(CMARB) were estimated from these data. Literature estimates for carcass weight
constant marbling score heritability values include 0.28, 0.33, and 0.73, respectively, from
Veseth et al. (1993), Cundiﬂ‘ et al. (1971), and Brackelsberg et al. (1971). The referenced
heritability estimates for CMARB indicate that the trait is moderate to highly heritable,
and this study’s results show a likewise conclusion.

This study estimated a heritability value of 0.38 i 0.080 for carcass weight
constant longissimus muscle area (CLMA). Heritability estimates from the literature for
CLMA heritability values include 0.32 and 0.40. Koots et al. ( 1994a) found 15 heritability

values for CLMA to average 0.41. The literature values suggest that CLMA is a

92

Table 24. Model 3 average phenotypic correlations for growth traits '.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Trait BW w w ADG EOTWT
Birth Weight

-4419?»th Weaning Weight 0.34 .................................................
Adjusted Yearling Weight 0.22 0.67
...Bgstrwsanins..Aysrassyailyﬁain 0-14 0-20 092.... .. ..............
End OfTest Weight 0.33 0.64 0.59 0.41

 

' BW = Birth Weight; W = Adjusted Weaning Weight; YW = Adjusted Yearling
Weight; ADG = Post-weaning Average Daily Gain; EOTWT = End of Test Weight.

93

Table 25. Model 3 average phenotypic correlations for carcass traits "b.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Trait CARCWT MARB LMA FAT CARCGR CUT
Marbling Score 0.04 .................................................................
Longissimus Muscle Area 0.51 0.11 .
ththbFatThrckness007 41.7 ............. 4.12 .................................................................................
Carcass 99.49 ................................................. 9:93 ...................... :0. .97 ............. 7.9.1.2. ............. 9.6.4..-- .............................................................
Cutability 0 32 0.17 0.82 -0.62 -0 46

 

Rib Fat Thickness; CARCGR = Carcass Grade; CUT = Cutability.
" Carcass traits were analyzed with an age at slaughter covariate.

 

94

 

 

 

Table 26.

Model 3 average phenotypic correlations betweergrowth and carcass traits “".
Trait BW WW YW ADG EO'I'W'I'

Carcass Weight 0.43 0.69 0.90 0.59 0.86

Marbling Score 0.08 0.04 0.03 0.02 0.03

 

Longissimus Muscle Area 0.24 0.38 0.42 0.23 0.40

 

 

12thRibFat Thickness -012” 0.03 0.09 0.07 0.09

CarcassGrade """"" V """" ' """" -0.10 -002 0.00“ 0.01 0.00

 

Cutability 0.25 0.25 0.25 0.131 0.23

 

 

 

 

 

 

 

' BW = Birth Weight; W = Adjusted Weaning Weight; YW = Adjusted Yearling Weight;
ADG = Post-weaning Average Daily Gain; EOTWT = End of Test Weight.

" Carcass traits were analyzed with an age at slaughter covariate.

95

32.22.2450 222N203 4482.8 4 222222 422422 443240 .

 

 

444.0 228.4 Sod owed H 44.0 435242222. 22242 452 22242
vovsm 34.3 420m Sod H 44.0 82 4282222 4:82mm2w223
3.4 242.4 444.4 2444 H 44.4 284 22222222242
43.42% 2.22. 2% 23.02 48.22 H 22.0 223243 wc§>> 2234.22.32
822222225 822422; 8224.5> 844m 222222222225 2242.4
42222226222222 24222422222022>=m 02202240 84222228222942
222222 42222222222842

 

.. 8222222228 3232.29» 4222542202222 22:22 2322.22.25» 22222202222222.4220 2322.22.29» 428.2% 22222222252422 .2 2022422 .44 0.22.2.2.

96

Table 28. Model 4 genetic, environmental, and phenotypic correlations “ ”.

 

 

Trait WW MARB LMA FAT

Adjusted Weaning Weight 0.11 -0. 11 -0.11 0.28

Marbling Score 0.07 0.28 0.22 -0.29
0.07

Longissimus Muscle Area 0.08 0.01 0.38 -0.29
0.08 0.10

12th Rib Fat Thickness -0.04 -0. 16 -0.26 0.36

 

 

 

-0.04 -0.16 -0.26
' Genetic correlations above diagonal, environmental correlations above the phenotypic
correlations below diagonal, heritability on diagonal.

b W = Adjusted Weaning Weight; MARB = Marbling Score; LMA = Longissimus
Muscle Area; FAT = 12th Rib Fat Thickness.

 

97

moderate to highly heritable trait, and this study’s CLMA heritability estimate is supported
by literature estimates.

Koots et al. (1994a) analyzed 15 heritability estimates for carcass weight constant
12th rib fat thickness (CFAT) and reported a mean value of 0.44. A review of the
literature revealed CFAT heritability of 0.43 and 0.53 from Brackelsberg et al. (1971), and
Cundiﬂ' et al. (1971), respectively. The current study’s CFAT heritability value of 0.36

i 0.080 indicates the CFAT heritability estimate is below the literature CFAT estimates.

4.4.2 Correlation Estimates

Genetic, environmental, and phenotypic correlations for Model 4 that used a
carcass weight covariate are included in Table 28. As was expected from the results of
Benyshek (1981), the genetic and phenotypic correlations between the trait combinations
that had either slaughter age or carcass weight modeled as a covariate did yield quite
similar results. The lone exception was the genetic and phenotypic correlation between
WW and LMA. The genetic correlation between WW and age constant longissimus
muscle area versus WW and carcass weight constant longissimus muscle area is 0.09 and
-0.11, respectively. Additionally, the phenotypic correlation between WW and age
constant longissimus muscle area versus WW and carcass weight constant longissimus
muscle area is 0.38 and 0.08, respectively. Moreover, the heritability values for age
constant longissimus muscle area and carcass weight constant longissimus muscle area

was 0.29 and 0.38, respectively. The diﬂ’erences can be explained due to the residual

98

variance being larger in these data when slaughter age is held constant versus a carcass

weight constant.

5. CONCLUSIONS

The results of this study indicate that there was no detectable selection bias in
these data, as the multiple trait heritability values for both grth and carcass traits closely
correspond to those achieved from the single trait analysis. Additionally, this study
concluded that heritability values for such grth traits such as weaning weight, yearling
weight, post-weaning average daily gain, and end of test weight are low to moderately
heritable. When selection pressure is placed upon these traits, some improvement will
result ﬁ'om the selection. The data indicated moderate heritability values for such carcass
traits as marbling score, longissimus muscle area, 12th rib fat thickness, carcass grade, and
cutability. These results suggest that a moderate response to selection can be achieved
when selecting for these carcass traits, and also that more response to selection can be
achieved from selecting for these carcass traits than for selecting for the grth traits
analyzed in this study.

Additionally, this study indicates that selection for faster growing, heavier muscled
cattle can be accomplished but not without antagonistic results. Yearling weight had both
high genetic and phenotypic correlations to other growth traits such as weaning weight,
post-weaning average daily gain, and end of test weight. Moreover, yearling weight was
highly correlated to carcass weight and had a -0. l6 correlation to marbling score, but
unfortunately the current data indicated a genetic correlation of 0.26 between yearling
weight and fat thickness in addition to a negative genetic correlation of yearling weight to

cutability. Marbling score had a numerically inverse genetic correlation to end of test

99

100

weight, however, phenotypically was positively correlated to end of test weight. An
additional antagonistic environmental correlation Was concluded ﬁom the correlation of
cutability and yearling weight, which had a positive phenotypic correlation of 0.25 but a
negative genetic correlation of -0.08. Nevertheless, selection for faster growing cattle
will result in heavier birth weights, along with lower cutability cattle with more fat. When
Canadian Charolais breeders select for heavier muscled, higher cutability cattle, they will
again see a moderate increase in birth weights in addition to a low to moderate decrease in
marbling. Therefore, producers must identify seedstock which do not follow these genetic
antagonisms to be able to produce beef that is proﬁtable and ﬁts into the industry’s

speciﬁcations to achieve consumer acceptance.

6. LITERATURE CITED

Arnold, J.W., J.K. Bertrand, L.L. Benyshek and C. Ludwig. 1991. Estimates of genetic
parameters for live animal ultrasound, actual carcass data, and grth traits in beef
cattle. J. Anim. Sci. 69:985.

Banks, B.D. 1986. Estimation of genetic parameters and sire rankings for Holstein linear
type scores and milk production by multiple trait analysis. Ph.D. Dissertation.
Michigan State University, East Lansing.

Barkhouse, KL. 1993. Genetic parameters for tenderness traits of ROS Indicus by 305
T bums Crosses. M.S. Dissertation. University of Nebraska, Lincoln.

Beef Improvement Federation. 1990. Guidelines for Uniform Beef Improvement
Programs (6th Ed). Oklahoma State University, Stillwater.

Benyshek, LL. 1981. Heritabilities for grth and carcass traits estimated from data on
Herefords under commercial conditions. J. Anim. Sci. 53:49.

Brackelsberg, P.O., EA. Kline, R.L. Willham and LN. Hazel. 1971. Genetic parameters
for selected beef-carcass traits. J. Anim. Sci. 33:13.

CCA. 1994. Charolais Sire Summary: Conception to Consumer Progeny Test Program
1992-1994. Canadian Charolais Association. Calgary, Alberta.

Cundiﬂ‘, L.V., K.E. Gregory, R.M. Koch and GE. Dickerson. 1971. Genetic relationships
among grth and carcass traits of beef cattle. J. Anim. Sci. 33:550.

Cunningham, BE. 1989. Investigation of sire by breed of dam interaction within

percentage Simmental groups for birth weight and 205D weight in the US.

101

102

Simmental population. Ph.D. Dissertation. Michigan State University, East
Lansing.

de Rose, ER 1992. Canadian EPD’s. Proc. Beef Improvement Federation Research
Symposium pp 5361. Portland, Oregon.

Dickerson, GE. 1958. Techniques for research in quantitative animal genetics.

American Society of Animal Production.

 

Falconer, D.S. 1960. Introduction to Quantitative Genetics. The Ronald Press
Company. New York, New York.

Graser, H.U., S.P. Smith and B. Tier. 1987. A derivative-free approach for estimating
variance components in animal models by restricted maximum likelihood. J. Anim.
Sci. 64: 1362.

Hartley, HO. and J .N.K. Rao. 1967. Maximum-likelihood estimation for the mixed
analysis of variance model. Biometrika 54:93.

Harville, DA. 1977. Maximum likelihood approaches to variance component estimation
and to related problems. J. Am. Stat. Assoc. 72:320.

Henderson, CR. 1953. Estimation of variance and covariance components. Biometrics
9:226.

Henderson, CR. 1973. Sire evaluation and genetic trends. In: Proc. Anim. Breed.
Genet. Symp. in Honor of J. L. Lush. p10. ASAS and ADSA, Charnpaign, IL.

Henderson, CR. 1974. Use of relationships among sires to increase accuracy of sire

evaluation. J. Dairy Sci. 5821731.

103

Henderson, CR. 1975. Comparison of alternative sire evaluation methods. J. Anim. Sci.
41:760.

Henderson, CR. 1976. A simple method for computing the inverse of a numerator
relationship matrix used in prediction of breeding values. Biometrics 32:69.
Henderson, CR. and R.L. Quaas. 1976. Multiple trait evaluation using relatives’ records.

J. Anim. Sci. 43:1188.

Jensen, J. and P. Madsen. 1993. A user’s guide to DMU: A package for analyzing
multivariate mixed models. National Institute of Animal Science Research Center
Foulurn, Denmark.

Johnson, D. and R. Thompson. 1995. Restricted maximum likelihood estimation of
variance components for univariate animal models using sparse matrix techniques
and average information. J. Dairy Sci. 78:449.

Johnston, D.J., L.L. Benyshek, J.K. Bertrand, M.H. Johnson and GM. Weiss. 1992.
Estimates of genetic parameters for growth and carcass traits in Charolais cattle.
Can. J. Anim. Sci. 72:493.

Jones, S.D.M. 1993. Grading effects on beef marketing. In: Proc. “Matching Beef
Cattle to Western Environments” Symposium of Western Regional Coordinating
Committee for Beef Cattle Research. pp 106-110. Calgary, Alberta.

Koch, R.M., L.V. Cundiﬁ‘ and KB. Gregory. 1982. Heritabilities and genetic,
environmental and phenotypic correlations of diverse biological types and their

implications in selection programs. J. Anim. Sci. 5521319.

104

Koots, K.R., J.P. Gibson, C. Smith and J.W. Wilton. 1994a. Analyses of published
genetic parameter estimates for beef production traits. l. Heritability. Animal
Breeding Abstracts 62:309.

Koots, K.R., J .P. Gibson, and J.W. Wilton. 1994b. Analyses of published genetic
parameter estimates for beef production traits. 1. Phenotypic and genetic
correlations. Animal Breeding Abstracts 62:825.

Madsen, P., J. Jensen and R. Thompson. 1994. Estimation of (co)variance components
by REML in multivariate mixed linear models using average of observed and
expected information. In: Proc. of the 5th World Congress on Genetics Applied
to Livestock Production 22: 19-22.

Marshall, D.M. 1994. Breed differences and genetic parameters for body composition
traits in beef cattle. J. Anim. Sci. 72:2745.

Ngwerume, F. 1994. Application of a multitrait animal model to predict next test-day
milk production. Ph.D. Dissertation. Michigan State University. East Lansing.

Nunez-Dominguez, R., L.D. Van Vleck, K.G. Boldman, and L.V. Cundiﬂ‘. 1993.
Correlations for genetic expression for growth of calves of Hereford and Angus
dams using a multivariate animal model. J. Anim. Sci. 71 :2330.

Patterson, H.D. and R. Thompson. 1971. Recovery of inter-block information when
block sizes are unequal. Biometrika 58:545.

Pollak, E]. and R.L. Quaas. 1980. Mixed model methodology for farm and ranch beef

cattle testing programs. J. Anim. Sci. 5121277.

105

Pollak, E.J., J. van der Werf, and RL. Quaas. 1984. Selection bias and multiple trait
evaluation. J. Dairy Sci. 67:1590.

Quaas, RL. 1976. Computing the diagonal elements and inverse of a large numerator
relationship matrix. Biometrics 32:949.

Robertson, A 1977. The eﬂ‘ect of selection on the estimation of genetic parameters. Z.
Tierz. Zuechtungsbiol. 94:131-135.

Saama, PM. 1992. Genetic parameters for partitioned uses of energy intake estimated
ﬁ'om ﬁeld collected and calorimetric data on the same lactating Holstein cows.
M.S. Dissertation. Michigan State University, East Lansing.

SAS®ISTAT Users guide. 1990. SAS Institute, Inc., Cray, NC.

Schaeﬂ‘er, LR 1984. Sire and cow evaluation under multiple trait models. J. Dairy Sci.
67: 1567.

Searle, SR. 1961. Phenotypic, genetic and environmental correlations. Biometrics
17:474.

Swiger, L.A., W.R. Harvey, D.O. Everson and KB. Gregory. 1964. The variance of
intraclass correlation involving groups with one observation. Biometrics 20:818.

Van Vleck,, L.D. 1968. Selection bias in estimation of the genetic correlation.
Biometrics 24:951.

Van Vleck, L.D., A.F. Hakim, L.V. Cundiﬂ', R.M. Koch, J.D. Crouse and K.G. Boldman.
1992. Estimated breeding values for meat characteristics of crossbred cattle with

an animal model. J. Anim. Sci. 70:363.

106

Veseth, D.A., W.L. Reynolds, J.J. Urick, T.C. Nelsen, RE. Short, and DD. Kress. 1993.
Paternal half-sib heritabilities and genetic, environmental, and phenotypic
correlation estimates from randomly selected Hereford cattle. J. Anim. Sci.
71 : 1730.

Walter, JP and IL. Mao. 1983. Consideration of mates in ranking Guernsey sires for
type. J. Dairy Sci. 66:2568.

Wilson, D.E., R.L. Willham, S.L. Northcutt, and G.H. Rouse. 1993. Genetic parameters
for carcass traits estimated ﬁom Angus ﬁeld records. J. Anim. Sci. 71 :2365.

Woodward, B.W., E.J. Pollak, and R.L. Quaas. 1992. Parameter estimation for carcass
traits including grth information of Simmental beef cattle using restricted

maximum likelihood with a multiple-trait model. J. Anim. Sci. 70: 1098.

1E5

 

"‘71111111111111111111

4