204 P
. Parkkonen et al. Livestock Production Science 64 2000 203 –213
predicts beef producing ability better than growth or animals according to the European cattle identifica-
live weight. In addition, using progeny results leads tion system as the first company in Finland. The
to improved reliability compared to individual mea- period of data collection covered 20 months from the
surements of young bulls. However, carcass data has beginning of January 1996 to the end of August
not been available for animal breeding because 1997. During that time more than 110 000 head of
slaughter houses have traditionally used their own cattle were slaughtered in the two participating
identification system, and it has not been possible to slaughter houses. Pedigrees were obtained from the
combine this system with other cattle registers. In the database of Agricultural Data Processing Centre
new European cattle identification system introduced including parents and grandparents for the slaug-
in Finland in the beginning of 1995 the same identity htered animals registered within milk recording
follows an animal from birth to carcass, thus en- system. The pedigree data set included over 180 000
abling environmental and pedigree information to be animals.
merged with carcass data. Sixty-three percent of slaughtered animals were
Carcass traits of cattle, e.g., slaughter weight, Finnish Ayrshires Ay and 26 Holstein-Friesians
fleshiness and fatness, have been studied considera- HFr. Other breeds and their crosses were too rare
bly, and most of the traits have been found to be of in the data set for estimation of genetic parameters,
high or moderate heritability Wilson et al., 1976; and it was also considered important to study carcass
Koch, 1978; Lamb et al., 1990; Robinson et al., traits in the most common breeds that produce the
1990; Arnold et al., 1991; Gregory et al., 1994; majority of beef. Thus, only purebred Ay or HFr
Wheeler et al., 1996. However, the results can not carcasses were included in the analyses.
be easily generalised into Finnish cattle population, Data was further limited to bulls and heifers that
because most of the studies have involved beef were slaughtered at the age of 300 through 899 days,
breeds which are only of marginal importance in with carcasses required to have slaughter weight of
Finland. Moreover, definitions of carcass traits and at least 130 kg. Cows were excluded from the
the models used in analyses differ in various coun- analyses in this study because cow information will
tries e.g., Jones et al., 1994; EU-ROP, 1995; Harris not be included in the possible carcass quality
et al., 1995; United States Department of Agricul- indices.
ture, 1997. The prerequisites for incorporating The data was divided in subsets to study whether
carcass traits into Finnish dairy cattle breeding the factors affecting carcass traits and their genetic
program are estimation of genetic parameters and parameters differ in different breeds and sexes. The
development of an appropriate evaluation model for primary subsets were Ay bulls AyB with 22 231,
carcass traits in Finnish cattle population. HFr bulls HFrB with 8711, Ay heifers AyH with
The aim of this study was to investigate the factors 5328 and HFr heifers HFrH with 1918 carcasses.
affecting carcass traits in Finnish Ayrshire and These subsets were analysed with all the models and
Holstein-Friesian, and to estimate heritabilities in methods used in this study. Combinations within
data sets divided by breed and sex. In addition, sexes and breeds were considered in combined
performance of animal model, sire model and sire subsets of Ay and HFr bulls AyHFrB, Ay and HFr
maternal grandsire model were compared in order to heifers AyHFrH, Ay bulls and heifers AyBH, and
find the model best suited for practical evaluation of HFr bulls and heifers HFrBH. Finally, all animals
carcass traits in Finnish dairy cattle breeding pro- were analysed together AyHFrBH. These combined
gram. data sets were analysed only with animal model
using univariate analysis. The traits studied were slaughter weight, and
2. Materials and methods carcass fleshiness and fatness. Slaughter weight is
measured within 2 h from slaughter, and it is the Data for analyses was collected from Northern and
weight of carcass without head, hide and abdominal Central Finland in two slaughter houses owned by
organs, minus 2 of hot carcass deduction. Fleshi- Lihakunta Oyj, which began to identify slaughtered
ness and fatness are judged subjectively according to
P . Parkkonen et al. Livestock Production Science 64 2000 203 –213
205
European Union SEUROP classification system EU- Sampling GS method with
MTGSAM
software Van ROP, 1995. In Finland, fleshiness is judged in 11
Tassell and Van Vleck, 1995. For comparison of classes: P 2 , P, P 1 , O 2 , O, O 1 , R 2 , R, R 1 , U
methods also sire models were analysed with GS and E, from worst to best respectively. Fatness is
method. judged in five classes numbered from 1 to 5, with
The following animal model was assumed in class 1 being the leanest and class 5 the fattest. In
analysing the within breed and sex data subsets AyB, this study, fleshiness was transformed to numbers so
HFrB, AyH and HFrH: that the classes from P 2 to R 1 were replaced by
y 5
m 1 slaughter house 1 year-month
ijklmn i
j
numbers from 1 to 9. Due to the lack of subclasses in U and E, they were numbered as 11 and 14,
1 age 1 c 1 a 1 ´
k l
m ijklmn
respectively. Figs. 1 and 2 illustrate frequency distributions of fleshiness and fatness in the data,
where y 5record of slaughter weight, fleshiness
ijklmn
respectively. or fatness,
m 5overall mean, slaughter house 5fixed
i
Data editing and preliminary analyses were done effect of ith slaughter house i 51,2, year-month 5
j
on
WSYS
and
WSYS
-
L
software Vilva, 1992; 1997. fixed effect of jth month of slaughter j 51–20,
For estimation of variance and covariance compo- age 5fixed effect of kth age class k51–14, c 5
k l
nents two methods were used. Animal models and random effect of lth herd, a 5random additive
m
sire models were solved by VCE4.0 software genetic effect of mth animal, and
´ 5random
ijklmn
Groeneveld, 1997 using Restricted Maximum residual effect. When analysing combined data sets
Likelihood REML method. Statistical significance AyHFrB and AyHFrH, also breed 5fixed effect of
o
of contrasts between different levels of fixed effects oth breed o 51,2 was included, and in data sets
in mixed models was tested by F-test in PEST AyBH and HFrBH sex 5fixed effect of pth sex
p
software Groeneveld, 1990. Sire maternal grandsire p 51,2 was included. When analysing all animals,
models could not be solved using VCE4.0 due to the data set AyHFrBH, both breed and sex were in-
1 ]
multiplier in the incidence matrix Z. Thus, esti-
cluded in the model in addition to the previous
2
mates of variance and covariance components from factors.
sire maternal grandsire models were solved by Gibbs There were two slaughter houses with 55 of
Fig. 1. Frequency distribution of fleshiness in bulls and heifers.
206 P
. Parkkonen et al. Livestock Production Science 64 2000 203 –213
Fig. 2. Frequency distribution of fatness in bulls and heifers.
carcasses coming from the bigger one. The original different traits i,i
9 51,2,3 and i ± i9 were assumed 20-month data collection period was retained in 20
to be covc ,c 5 I s
, vara , a 5 A s
and
i i 0
ci,i 9 i
i 9 ai, i 9
classes for year-month of slaughter because there cov´ ,´ 5 I
s .
i i 9
´i,i 9
was no logical connection between the consecutive Heritability was estimated as the proportion of the
2
months or the same months in different years. Rather additive genetic variance of the total variance, h 5
2 2
2 2
2
than using age at slaughter as covariate it was s s 1 s 1 s . Within herd heritabilities h
a a
c ´
w 2
2 2
2
classified in 14 classes 1510, 11 and 12 mo, 2513 were estimated as h 5
s s 1 s .
w a
a ´
mo, 3514 mo, . . . , 12523 mo, 13524 and 25 mo In sire model, the genetic effect of an animal was
and 14526 to 30 mo of age. substituted by the genetic effect of a sire, and in sire
Slaughtered animals originated from 6740 herds, maternal grandsire model the same section was
1
with a quarter of herds having only one observation
]
substituted by the genetic effect of the sire and of
2
in the data. Dividing data in subsets further increased the genetic effect of the maternal grandsire.
the proportion of small herds in the data subsets. The All observations were kept in analyses when using
herds with few carcasses could not be left out the animal model. With the sire model, only sires
without losing a considerable amount of information, with five or more progeny at the data set were
so it was decided to keep all the herds in analyses as included, and with sire maternal grandsire model a
a random sample of herds in the area. sire or a maternal grandsire was accepted only if it
The distributions of random effects were assumed existed in a pedigree of at least two slaughtered
multivariate normal with zero means and varc 5 animals. The restrictions decreased the number of
2 2
2
I s , vara 5 As , and var´ 5 Is . When using
observations but even more they decreased the
c a
´
multitrait models, the expected values of random number of sires thus increasing the number of
effects and the covariances between them were progeny per sire Table 1. The effect of restrictions
assumed zero. The variance of each random effect on parameter estimates was studied by comparing the
for the three traits i 51,2,3 was assumed to be solutions obtained using animal model both for the
2 2
2
v
arc 5 I s
, vara 5 A s
and var´ 5 I s
. unlimited data sets and the data sets limited for sire
i ci,i
i ai,i
i ´i,i
The covariances between the random effects in maternal grandsire model.
P . Parkkonen et al. Livestock Production Science 64 2000 203 –213
207 Table 1
Number of slaughtered animals N and sires S in different models and data subsets Breed,
Animal Sire
Sire maternal
a
sex model
model grandsire model
N S
N S
N S
AyB 22 231
892 21 518
366 21 273
478 HFrB
8711 361
8440 180
8305 226
AyH 5328
558 4904
310 4815
400 HFrH
1918 248
1688 126
1697 177
a
AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyH–Ayrshire heifers; HFrH–Holstein-Friesian heifers.
The results from different methods were compared 3. Results
by solving sire models both by REML and GS methods. When using GS method, the slaughter
The average slaughter weight for bulls was 273 kg house, the year-month of slaughter and the age at
and for heifers 203 kg Table 2. Carcasses of HFr slaughter were given flat prior distributions.
MTGSAM
bulls were on average 10 kg heavier than carcasses software provides inverted Wishart distribution for
of Ay bulls, while in heifers the difference between the prior distribution of variance and covariance
breeds was 8 kg. The average fleshiness of all components. Starting values were derived from the
carcasses was 4.3, i.e., between classes O2 and O. models solved by REML method. The convergence
Thus an average carcass had profiles from straight to criterion of Gauss–Seidel iteration was 0.0001. GS
concave, and average muscle development EU-ROP, algorithm was repeated for 30 000 rounds saving the
1995. Carcasses of bulls were classified on average solutions of every 30th round. Thus, the sample size
one grade better than carcasses of heifers, and HFr in point estimation was 934.
was classified 0.3 grades better than Ay. The average
Table 2 Number of observations N , means x, standard deviations s, coefficients of variation V , and minimum Min and maximum Max
values of studied traits in different data subsets
a
Trait Breed, sex N
x s
V Min
Max Slaughter weight
, kg AyB
22 231 270
40.8 15.1
130 466.0
HFrB 8711
280 41.6
14.9 131
490.5 AyH
5328 201
35.8 17.8
130 412.5
HFrH 1918
209 37.5
17.9 130
354.5 Fleshiness
AyB 22 231
4.43 0.99
22.4 11
HFrB 8711
4.75 1.02
21.4 11
AyH 5328
3.50 0.93
26.6 8
HFrH 1918
3.85 1.03
26.8 14
Fatness AyB
22 231 2.15
0.41 18.8
5 HFrB
8711 2.19
0.45 20.5
1 5
AyH 5328
2.69 0.79
29.3 5
HFrH 1918
2.81 0.89
31.7 5
a
AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyH–Ayrshire heifers; HFrH–Holstein-Friesian heifers.
208 P
. Parkkonen et al. Livestock Production Science 64 2000 203 –213
fatness of all carcasses was 2.27. In class 2, carcas- estimated to be low, at most 0.21, in AyB, HFrH and
ses are slightly fat covered with flesh visible almost AyH. Corresponding estimates from HFrH were
everywhere EU-ROP, 1995. Carcasses of heifers outside this range; however, the structure of HFrH
were on average 0.5 grades fatter than carcasses of data set was poor due to the large number of herds
bulls. HFr heifers were also fatter than Ay heifers, and sires compared to the small number of records.
but there was no difference between breeds in bulls. All within herd correlations were high, especially
Slaughter weights as well as fleshiness and fatness the correlations between slaughter weight and fleshi-
grades tended to decrease during the 20 month ness or fatness that were up to 0.73–0.93 depending
observation period. However, the trend was neither on the data set Table 5. There was little difference
linear nor similar in different data subsets. The between phenotypic and environmental correlations,
differences between year-months were statistically of which the correlation between slaughter weight
significant P .0.05 in all data subsets. The average and fleshiness was the highest 0.53–0.74, the
ages at slaughter varied between months as well. exception being again the data set HFrH.
On average, the animals were slaughtered at the Restrictions on data for sire and sire maternal
age of 18.5 months. Heifers were 1.5 months older grandsire models removed records with little or no
than bulls at slaughter, and HFr animals were connection with other records. Restrictions did not,
slaughtered 0.5 months younger than Ay animals. however, affect estimates of heritability by more
Slaughter weight increased from the youngest to the than 0.01 units when differently restricted HFrB and
oldest age class by 115 kg in bulls and by 90 kg in AyH datasets were analysed with animal model.
heifers. Fleshiness and fatness increased also with Thus results from different models are comparable,
age. Improvement of fleshiness was faster in bulls although the data sets used in the analyses were not
but gain of fat was faster in heifers. In most data exactly the same. The results for sire models that
subsets, however, the heaviest carcasses with the were obtained either with REML or GS methods
highest fleshiness and fatness grades were not in the from the same data sets did not differ from each
oldest age class. other either, showing that results by different meth-
Estimates of heritability for slaughter weight were ods agree with each other Table 6. No matter what
relatively low in all data subsets, varying from 0.07 model
or method
was used,
the estimated
to 0.14 Table 3. Respective estimates of within heritabilities and fractions of total variance due to
herd heritability were somewhat higher, from 0.15 to herds in the three biggest data sets were almost the
0.29. Heritabilities in HFr data sets were lower than same, differences being in the bounds of standard
heritabilities in Ay data sets. Variation between herds errors of the estimates.
caused approximately one half of the total variance in slaughter weight.
Fleshiness was estimated to have heritability of 4. Discussion
0.16–0.31, within herd heritability of 0.29–0.39 and the fraction of total variance due to herds of 0.20–
The data used in this study represented well the 0.26, depending on the data set used in the analyses
overall carcass quality of all the bulls and heifers Table 3. Estimates of heritability for fatness were
slaughtered in Finnish slaughter houses in 1996 about the same magnitude as for slaughter weight,
TIKE, 1997. Only in fleshiness carcasses in this varying between 0.08 and 0.16 in different data sets.
study were 0.20 units poorer than the average in the Estimated within herd heritabilities for the trait were
whole country. This difference was probably due to from 0.12 to 0.29, and the herds caused about 23–
the exclusion of beef breeds from the data set in this 47 of the total variance of fatness.
study. Estimates of genetic correlation between slaughter
Demand for beef in Finland varied somewhat weight and fleshiness were from 0.65 to 0.66 in
during the data collection period. This caused uneven AyB, AyH and HFrH, and 0.38 in HFrB Table 4.
numbers of animals to be slaughtered in different Genetic correlations between slaughter weight and
months and the average age of animals at slaughter fatness, and fleshiness and fatness, instead, were
to fluctuate. At the highest month, 3254 animals
P . Parkkonen et al. Livestock Production Science 64 2000 203 –213
209 Table 3
2
Number of records N , number of herds n, estimates of heritability h and their standard errors se , estimates of within herd
h 2 2
2
heritability h , and herd effects c and their standard errors se for studied traits from univariate analyses with animal model
w c 2
a 2
2 2
Trait Breed, sex N
n h 6se
h c 6se
h 2 w
c 2
Slaughter weight AyB
22 231 4381
0.1360.01 0.26
0.5260.01 HFrB
8711 2957
0.0960.01 0.19
0.5360.01 AyHFrB
30 942 5140
0.1160.01 0.24
0.5360.01 AyH
5328 2903
0.1460.02 0.29
0.5260.01 HFrH
1918 1232
0.1060.04 0.23
0.5760.02 AyHFrH
7246 3597
0.1460.02 0.28
0.5160.01 AyBH
27 559 5797
0.1260.01 0.23
0.4760.01 HFrBH
10 629 3570
0.0760.01 0.15
0.5060.01 AyHFrBH
38 188 6740
0.1160.01 0.21
0.4860.01 Fleshiness
AyB 22 225
4381 0.1760.01
0.22 0.2460.01
HFrB 8709
2956 0.2260.02
0.28 0.2260.01
AyHFrB 30 934
5140 0.1860.01
0.24 0.2460.01
AyH 5328
2903 0.1760.02
0.23 0.2660.01
HFrH 1915
1231 0.3160.05
0.39 0.2060.02
AyHFrH 7242
3596 0.2160.02
0.27 0.2560.01
AyBH 27 552
5797 0.1660.01
0.20 0.2360.01
HFrBH 10 624
3569 0.2160.02
0.26 0.2060.01
AyHFrBH 38 176
6740 0.1760.01
0.21 0.2260.01
Fatness AyB
22 225 4381
0.1260.01 0.16
0.2360.01 HFrB
8711 2957
0.0860.01 0.12
0.2960.01 AyHFrB
30 936 5140
0.1060.01 0.14
0.2660.01 AyH
5328 2903
0.1460.02 0.23
0.3760.01 HFrH
1916 1231
0.1660.04 0.29
0.4760.02 AyHFrH
7243 3596
0.1460.02 0.23
0.3860.01 AyBH
27 552 5797
0.1260.01 0.18
0.3060.01 HFrBH
10 627 3570
0.1060.02 0.15
0.3560.01 AyHFrBH
38 179 6740
0.1060.01 0.15
0.3360.01
a
AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyHFrB–Ayrshire and Holstein-Friesian bulls; AyH–Ayrshire heifers; HFrH– Holstein-Friesian heifers; AyHFrH–Ayrshire and Holstein-Friesian heifers; AyBH–Ayrshire bulls and heifers; HFrBH–Holstein-Friesian
bulls and heifers; AyHFrBH–Ayrshire and Holstein-Friesian bulls and heifers.
were slaughtered in the two participating slaughter Traditionally the carcass quality has been better in
houses, whereas at the lowest month the number was Finnish Friesian than in Finnish Ayrshire, but im-
only 671. The maximum difference between monthly portation of Holstein to the Finnish Friesian popula-
average ages at slaughter was 33 days. Since in some tion has weakened carcass quality of Finnish Hol-
months it was not possible to get all animals stein-Friesian
compared to
Ayrshire Liinamo,
slaughtered at the planned age, part of the animals 1997. In this study, percentage of Holstein genes
grew over aimed finishing point and gained fat. A among the slaughtered animals was not taken into
significant part of overaged animals may however account, but in the HFr data, carcasses were still
have been put on restricted feeding, as they neither heavier and had higher grades in fleshiness and
gained fat nor improved in fleshiness scores. Pos- fatness than Ay carcasses. Breed differences in
sibly this is why neither the heaviest carcasses nor fleshiness and fatness were larger in heifers than in
the highest average fleshiness and fatness grades bulls. Heifers were more susceptible to gain fat with
were in the oldest age class. age, and differences in carcass quality between sexes
210 P
. Parkkonen et al. Livestock Production Science 64 2000 203 –213 Table 4
Table 5 Estimates of genetic parameters for studied traits from multitrait
Fractions of total variance due to herds on diagonal with standard
a
analysis with animal model errors, within herd correlations above diagonal with standard
errors, and environmental correlations below diagonal estimated
b
Breed, sex Trait Trait
by multitrait analysis with animal model AyB
1 2
3
a
Breed, sex Trait Trait
1 Slaughter weight 0.1360.01
0.6660.03 0.1660.01
2 Fleshiness 0.65
0.1760.01 0.2160.05
AyB 1
2 3
3 Fatness 0.36
0.27 0.1260.01
1 Slaughter weight 0.5260.01
0.8660.01 0.7460.01
2 Fleshiness 0.65
0.2460.01 0.6360.01
HFrB 1
2 3
3 Fatness 0.38
0.28 0.2360.01
1 Slaughter weight 0.0960.01
0.3860.06 0.0560.09
2 Fleshiness 0.57
0.2060.02 0.1260.04
HFrB 1
2 3
3 Fatness 0.42
0.26 0.0860.01
1 Slaughter weight 0.5460.01
0.8260.01 0.7360.01
2 Fleshiness 0.61
0.2460.01 0.5560.02
AyH 1
2 3
3 Fatness 0.46
0.29 0.3060.01
1 Slaughter weight 0.1360.01
0.6660.06 20.0160.11
2 Fleshiness 0.61
0.1760.02 0.1860.07
AyH 1
2 3
3 Fatness 0.59
0.40 0.1360.01
1 Slaughter weight 0.5260.01
0.8160.01 0.8860.01
2 Fleshiness 0.74
0.2660.01 0.6960.02
HFrH 1
2 3
3 Fatness 0.56
0.44 0.3760.01
1 Slaughter weight 0.1160.03
0.6560.07 0.7660.07
2 Fleshiness 0.54
0.3160.04 0.4460.10
HFrH 1
2 3
3 Fatness 0.68
0.39 0.2060.04
1 Slaughter weight 0.5860.02
0.7360.04 0.9360.01
2 Fleshiness 0.53
0.2260.02 0.6960.04
a
Heritabilities on diagonal with standard errors, genetic correla- 3 Fatness
0.68 0.38
0.4860.02 tions above diagonal with standard errors, and phenotypic correla-
a
tions below diagonal. AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyH–
b
AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyH– Ayrshire heifers; HFrH–Holstein-Friesian heifers.
Ayrshire heifers; HFrH–Holstein-Friesian heifers.
estimated heritabilities for slaughter weight and increased with age. The different development of
fleshiness and fatness measured in SEUROP-scores sexes is in agreement with earlier reports e.g., Berg
have been 0.2260.03, 0.2360.03 and 0.2960.03, and Butterfield, 1976.
respectively de Jong, 1997, and 0.25, 0.26 and Estimated heritabilities, especially for slaughter
0.30, respectively Van der Werf et al., 1998. weight and fatness, were relatively low. Estimated
Estimated correlations revealed a positive genetic within herd heritabilities, however, were in range of
connection between slaughter weight and fleshiness. heritabilities estimated in previous studies, in which
Fatness was not genetically connected with other the herd has been taken as a fixed effect. Main
carcass traits. These correlations are favourable for emphasis in studying carcass traits has been on beef
work towards the breeding goal of carcasses with breeds. In Hereford and some other beef breeds,
high fleshiness and low fatness that give the type of heritability of carcass traits has been estimated to be
beef favoured by consumers at the present. Genetic moderate Wilson et al., 1976; Koch, 1978; Lamb et
correlation between fleshiness and slaughter weight al., 1990; Robinson et al., 1990; Arnold et al., 1991;
may however be somewhat overestimated, for large Gregory et al., 1994; Wheeler et al., 1996. In dairy
carcasses appear more muscular than small carcasses ¨
breeds, Kenttamies 1983 estimated heritabilities for and may thus unintentionally grade better than small
slaughter weight in Finnish Ayrshire bulls as carcasses of corresponding quality. Both the pheno-
0.2360.09 and in Friesian bulls as 0.6160.18. In the typic and environmental correlations and especially
same study, estimates for fleshiness were 0.1460.08 the correlations within herds show that fatness tends
in Ayrshire and 0.1560.13 in Friesian, and for to increase with slaughter weight and fleshiness.
fatness 0.0660.18 and 0.2660.15 in Ayrshire and Hence management and feeding seem to have a key
Friesian, respectively. In larger data sets of Dutch position in the production of animals of high carcass
Black and White and Dutch Red and White bulls, quality. Carcass quality can, however, be altered by
P . Parkkonen et al. Livestock Production Science 64 2000 203 –213
211 TABLE 6
of data sets in multitrait analyses, other alternatives
Heritabilities of studied traits from different types of multitrait
were also considered. Sire model is the simplest
a
models
model for estimating breeding values for sires, if the
b
Trait Breed, sex Model 1
Model 2a 2b Model 3
only group of animals for which the breeding values
Slaughter weight
for carcass traits need to be estimated is sires
AyB –
0.12 0.12 0.12
themselves. However, there are also relationships on
HFrB 0.09
0.08 0.08 0.08
maternal side through common grandsires, and those
AyH 0.13
0.11 0.09 0.13
can not be taken into account by sire model. In that
HFrH 0.11
0.00 0.00 0.05
aspect, a better option could be the use of a sire
Fleshiness
maternal grandsire model. However, in this study
AyB –
0.20 0.21 0.20
there were only minor differences in estimated
HFrB 0.20
0.25 0.25 0.23
variance and covariance components or their pro-
AyH 0.17
0.17 0.16 0.15
portions between different models. For the sake of
HFrH 0.31
0.21 0.20 0.27
computational resources, sire or sire maternal gran-
Fatness
dsire models are often preferred to animal models,
AyB –
0.14 0.14 0.15
and sire maternal grandsire models, in turn, are
HFrB 0.08
0.08 0.07 0.10
preferred to sire models as they include the sire path
AyH 0.13
0.12 0.10 0.13
of maternal pedigree. Nevertheless, estimated breed-
HFrH 0.20
0.11 0.10 0.12
ing values for carcass traits in cows may also be of
a
Model 1: Animal model, REML method three traits in AyB
interest in herd level. For that reason animal model,
could not be solved by animal model due to limits in computation-
whenever it is computationally feasible, might be the
al resources. Model 2a: Sire model, REML method. Model 2b:
most suitable model for practical evaluation of
Sire model, GS method. Model 3: Sire–maternal grandsire model, GS method.
carcass traits.
b
AyB–Ayrshire bulls; HFrB–Holstein-Friesian bulls; AyH–
The type and size of data in this study gave
Ayrshire heifers; HFrH–Holstein-Friesian heifers.
reliable estimates for genetic parameters, and might therefore be suitable also for the estimation of the
management changes only within the limits of the breeding values. Obtaining estimates of breeding
genetic potential of the animal which, in turn, can be value for carcass quality traits for young AI-bulls
further improved by breeding. does not lengthen generation interval, because the
When estimating breeding values for carcass traits beef producing progeny are slaughtered already
of Ay and HFr young bulls it might be feasible to before the milk producing daughters complete their
analyse the carcass data of their offspring for both first lactation. Moreover, Liinamo and van Arendonk
breeds and sexes together, even though there were 1998 have shown that genetic improvement in
some differences in genetic parameters between carcass traits does not retard genetic response in milk
breeds and sexes. Combining the two data sets might production traits. Thus breeding for carcass quality
reduce the number of herds with one or few carcas- in dairy cattle seems a quite feasible option for
ses, and thus improve the structure of data. However, improving the overall economy of cattle producing
in this study this effect was not strong when breeds sector.
were combined, as the majority of herds represented Nevertheless, there were some inadequacies in the
only one breed. On the other hand, heifers comprise data, one of them being the short time period of
only 20 of carcasses, and carcass quality estima- observations. As Finnish dairy herds in milk record-
tion could be solely based on bulls. Combining sexes ing average about 15 cows FABA, 1996, there was
may, however, be of value in predicting fatness more consequently a large amount of herds with only one
accurately, as heifers provide additional information or few carcasses in the data. An even more serious
on fatness while most of the bulls were graded as 2. defect was that the data did not cover whole country.
Animal model was fitted to the data first because Regional differences can somewhat be eliminated
of its property to take all the relationships into with statistical model used, but data collected from
account. Since there were problems due to the sizes an area covering the country more widely would
212 P
. Parkkonen et al. Livestock Production Science 64 2000 203 –213 1995. Relationship between USDA and Japanese beef grades.
represent all the young AI-bulls used better, and
Meat Sci. 39, 87–95.
provide them more slaughtered progeny than a
Hietanen, H., Ojala, M., 1995. Factors affecting body weight and
regional data. In fact, the bulls with large numbers of
its association with milk production traits in Finnish Ayrshire
progeny were popular progeny tested bulls, and some
and Friesian cows. Acta Agric. Scand. 45, 17–25. INTERBULL, 1996. Sire evaluation procedures for non-dairy-
of the young bulls had only few progeny in the data.
production and growth and beef production traits practiced in
Thus it is not feasible to estimate breeding values for
various countries. Bull. no. 13, Int. Bull Eval. Serv., Uppsala,
young AI-bulls until there is data available from
Sweden, 201 pp.
more slaughter houses.
Jones, S.D.M., Thorlakson, B., Robertson, W.M., 1994. The effect of breed type on beef carcass characteristics and Canadian
carcass grade. Can. J. Anim. Sci. 74, 149–151. de Jong, G., 1997. Genetische parameters voor slachtkenmerken
5. Conclusions