- Email: [email protected]
Estimated parameters of performance in jumping and dressage competition of the Dutch Warmblood horse
Livestock Production Science, 21 (1989) 333-345
333
Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands
Estimated Parameters of Performance in Jumping and Dressage Competition of the Dutch Warmblood Horse H.A. HUIZINGA 1 and G.J.W. VAN DER MEIJ 2
1Department of Animal Breeding, Wageningen Agricultural University, P.O. Box 338, 6700 AH Wageningen (The Netherlands) 2Department of Herd Health and Reproduction, Faculty o[ Veterinary Medicine, University of Utrecht, P.O. Box 80.151, Utrecht (The Netherlands) (Accepted for publication 28 November 1988)
ABSTRACT Huizinga, H.A. and van der Meij, G.J.W., 1989. Estimated parameters of performance in jumping and dressage competition of the Dutch Warmblood horse. Livest. Prod. Sci., 21: 333-345. The objective of this study is to estimate several genetic parameters in the Dutch Warmblood riding horse population. The traits involved are performances in jumping and dressage competition. The following parameters are estimated: heritabilities for jumping and dressage; phenotypic and genetic correlations between jumping and dressage; and phenotypic and genetic correlations between performances at different ages. These parameters are estimated by restricted maximum likelihood (REML). Data are from 6899 horses with performances in jumping and 10 408 horses with performances in dressage competition. The horses are sired by 205 and 237 stallions for the two traits, respectively. The progeny range in age from 4 to 8 years old. The performance trait is a cumulatively derived score, that reflects the level of performance in competition. A square root transformation of the score is most appropriate to normalize the data. For estimation of phenotypic and genetic parameters the data is split into two data sets according to the age of the sires {offspring sired by older vs. younger stallions). For estimating correlations between performances at 4, 5 and 6 years of age, performances of the offspring out of previous years are linked to the data. The most unbiased estimates of heritability for jumping and dressage are from data derived from the youngest offspring sired by the younger stallions and are 0.20 and 0.10, respectively. Genetic correlation between jumping and dressage ranges from -0.27 to 0.10. The phenotypic correlation between these traits ranges from 0.15 to 0.26. Phenotypic and genetic correlations between performances at 4, 5 and 6 years average 0.95 and 0.75, respectively. These latter results have important implications for genetic evaluation of breeding candidates in the population.
INTRODUCTION The active breeding population of the Dutch riding horse consists of about 9 0 0 0 m a t e d m a r e s y e a r - 1 w i t h s e r v i c e s f r o m a b o u t 150 s t a l l i o n s ( I n t h e S t r e n -
0301-6226/89/$03.50
© 1989 Elsevier Science Publishers B.V.
334 gen, 1987). Horses are registered by the Royal Warmblood Studbook of the Netherlands (KWPN). To predict the response to selection, to evaluate various breeding plans and to predict breeding values of breeding candidates for selection, a knowledge of genetic parameters is essential. The first phase in developing a breeding program is defining the breeding goal. Even when evaluation is restricted to performance traits, we still have to deal with a least two traits; jumping and dressage. Estimates of heritability of these traits in the Dutch Warmblood horse population are needed. Breeders and riders are frequently striving for simultaneous improvement in both traits. Before considering the combination of jumping and dressage in the breeding goal, knowledge of phenotypic and genetic (co)variances is of interest. Progeny testing in general will increase generation interval. To limit this disadvantage, e.g. stallions can be evaluated on the basis of performances of their first offspring at an early age. For justifying this, knowledge is needed of genetic correlations between performances at younger and older ages of the horse. Summarizing, there are three objectives in this paper: (1) to estimate heritability for jumping and dressage; (2) to estimate relationships between jumping and dressage; (3) to estimate relationships between the level of performance at different ages. REVIEW OF LITERATURE In France, earnings per year (Langlois, 1980; Tavernier, 1986) and in the F.R.G. earnings per start (Bruns, 1981; Meinardus and Bruns, 1987) in competition are the traits that are involved in the genetic evaluation of breeding candidates in the population. In Sweden, the number of placings and the number of starts are the criteria (Philipsson, 1975). Most procedures adjust for age, sex and year of performance. Riding club effects, which are expected in The Netherlands, were not taken into account in these studies. A review of estimates of heritability for jumping and dressage is given in Table 1. Estimates of heritability are in most cases higher for jumping than for dressage. However, Bruns (1981) found the opposite. Estimates of phenotypic correlation under field conditions assessed by Thafvelin et al. (1980) between gaits under saddle and jumping ability ranged from 0.23 to 0.30. Phenotypic correlations, which can be considered as repeatabilities, are reported by a number of studies, as shown in Table 1. Bruns (1981) and Meinardus and Bruns (1987) estimated genetic correlations between jumping and dressage indirectly by computing correlations between breeding values for the two traits. Estimates were up to 0.10 and 0.14, respectively. Bruns et al. ( 1985 ) estimated parameters from data from central performance tests of stallions. The estimates for genetic correlation between
335 TABLE 1 Estimates of heritabilities and repeatabilities for jumping and dressage in France, F.R.G. and Sweden Heritabilities
Langlois (1980) Tavernier (1986) Bruns (1981) Meinardus and Bruns (1987) Philipsson (1975)
Repeatabilities
Jumping
Dressage
Jumping
Dressage
0.15-0.19 0.20-0.25 0.13-0.14 0.18-0.21 0.16-0.29
-0.12-0.05
0.46-0.53 0.45 0.21-0.34
0.46-0.54 0.29-0.39
0.42-0.48
0.45-0.48
0.15-0.27 0.16-0.18 0.04-0.06
show jumping and gaits averaged 0.37. Their estimates for genetic correlation between free jumping and gaits averaged - 0.53 and between free jumping and riding ability - 0.06. Tavernier (1986) estimated genetic correlations between performances in jumping competition from 4 through 10 years of age. The genetic correlations averaged 0.75 between 4 and the other ages, 0.85 between 5 and the other ages and 1 between the remaining age classes. MATERIALS AND METHODS
For estimation of phenotypic and genetic parameters two datasets, A and B, are created. Dataset A is based on competition results from 1986. In these data, heritabilities for jumping and dressage, and phenotypic and genetic correlations between jumping and dressage are estimated. Dataset B is based on horses with competition results in 1986, while repeated observations from previous years are linked to this dataset. These repeated measures of the performances at an earlier age enhance the number of records considerably, while the number of horses involved did not change. These data are used for estimating phenotypic and genetic correlations between performances at different ages of the horse. Records from offspring over 8 years old in 1986 and data of stallions with less than 5 offspring are excluded. The remaining data sets contain 6899 horses in jumping competition sired by 204 stallions and 10 408 horses in dressage competition sired by 237 stallions. Table 2 shows the number of records year- 1 in the data sets. The trait which expresses performance in sport competition is the so-called "highest level" in sport, accomplished during the lifetime of the horse. The lifetime total is cumulative and reflects the level of performance of the horse
336 TABLE 2 Number of records per year in the data Number ofhorses in 1986
Jumping Dressage Content of Datasets: A
6899 10408
B
X
Number of repeated measures in: 1985
1984
1983
4696 7641
2628 4608
1107 2228
X
X
X
X
TABLE 3 Means and standard deviations for the traits in the two datasets Jumping
Dressage
Younger
Older
Younger
Older
Dataset A LT4-LT6 LT4-LT7 LT4-LT8
2.57 _+1.05 2.61 +- 1.07 2.67 +_1.09
2.65 _+1.05 2.74+_ 1.07 2.80 +_1.10
2.21 +- 1.16 2.26+_ 1.17 2.30 +_1.18
2.18 +- 1.16 2.28+_ 1.18 2.35 +_1.19
Dataset B LT4 LT5 LT6 LT7 LT8
2.17 +_0.93 2.54 +_1.03 2.75 +- 1.09 2.86+_ 1.12 2.99 +- 1.14
2.19 +-0.92 2.59 +_1.01 2.82 +- 1.10 2.95 +_1.11 3.02 +- 1.15
1.79 +_1.12 2.17 +_1.14 2.38+_ 1.17 2.47 +_1.19 2.58 +- 1.21
1.78 +- 1.12 2.13 +_1.15 2.36+_ 1.16 2.50_+ 1.19 2.58 +_1.23
in a discipline of competition. These scores, consisting of a character and number combination, are transformed to a linear scale. For a more detailed descript i o n o f t h e t r a n s f o r m a t i o n o f s c o r e s , see H u i z i n g a ( 1 9 8 7 ) . T o a c c o m p l i s h a m o r e n o r m a l l y d i s t r i b u t e d r e s i d u a l e r r o r t e r m , a s q u a r e r o o t t r a n s f o r m a t i o n is applied. I t is e x p e c t e d t h a t d a t a f r o m t h e o l d e r s t a l l i o n s h a v e b e e n s u b j e c t e d t o m o r e selection. Therefore, data are split into two groups according to birth date of t h e s t a l l i o n ( b e f o r e 1972 a n d 1972 o r a f t e r ) . M e a n s a n d s t a n d a r d d e v i a t i o n s f o r t h e t r a i t s i n t h e d i f f e r e n t d a t a s e t s a r e g i v e n i n T a b l e 3. From 4 to 5 years of age about 85% of the horses make progress during a c o m p e t i t i o n y e a r . F r o m 7 t o 8 y e a r s o f a g e o n l y a b o u t 35% a r e a b l e t o r e a l i z e p r o g r e s s . W h e n a h i g h l e v e l o f p e r f o r m a n c e is r e a c h e d , m a k i n g p r o g r e s s b e comes more difficult. But also factors related to breeding, export and culling
337 TABLE 4 Number of records and sires per dataset per trait Jumping
Dressage
Younger
Older
Younger
Older
DatasetA LT4-LT6 LT4-LT7 LT4-LT8
1563/ 94 2031/100 2396/101
2271/ 91 3479/100 4504/104
2536/103 3226/111 3778/114
3403/ 99 5106/115 6632/123
DatasetB LT4 LT5 LT6 LT7 LT8
992/ 1472/ 1312/ 800/ 364/
1549/ 97 2893/103 2804/100 2132/ 96 1025/ 90
2074/110 2560/105 2035/ 88 1206/ 69 550/ 60
2962/110 4649/122 4270/119 3092/115 1526/106
95 95 90 64 52
play a role, and for these reasons the datasets are split according to the age of the offspring. Only for Dataset A is this of importance and leads to three groups of records referring to lifetime totals (LT's) based on age: lifetime totals from 4 through 6 years; lifetime totals from 4 through 7 years; lifetime totals from 4 through 8 years. Dataset B is analysed within age classes. The respective lifetime totals are at 4 ( LT4 ), 5 (LT5) and 6 ( LT6 ) years of age. A description of the data results in frequencies as shown in Table 4. The following linear models are applied for analyses of the performance records of jumping and dressage. For Dataset A: Y~jt~,, = district~ + agei + sexl + sire~ + e,~l~o
For Dataset B: Y;kt.... = district~ + NYSk + sext + yearm + sire~ + eikl~no where: = observed score on the oth offspring; Y~jtno,, Yim . . . . District,. = effect of the ith district ( i = 1 .... 200); Agej = e f f e c t o f t h e j t h age-class (1"=4, 5, 6,); NYS~ ---effect of the number of years in competition (k--1, 2, 3); ---effect of the year ( m = 1983, 1984, 1985, 1986); Yearm Sex/ = effect of t h e / t h sex (l = stallions, mares, geldings); = random effect of the nth sire; Siren - r a n d o m residual term. e ijlno'e iklmno In the combined data of the younger and older stallions, 780 riding clubs for
338
jumping and 920 riding clubs for dressage are represented. For jumping as well as for dressage 62.8% of the riding clubs have less then 11 combinations participating in competition. By dividing the data according to age of the sires, the number of observations per riding club is further decreased. To increase these numbers, the number of riding clubs is reduced to 200 districts. Geographical and data similarities are used to join the riding clubs. Statistical methods generally used in the analysis of animal breeding assume that these data are a random sample from the population. However, this is often not true, also not for competition data from horses. From 137 traced stallions, covering 75% of the data with their offspring, only about 30% of the registered foals are brought into dressage competition and 20% into the jumping competition. Variance and covariance components are estimated by restricted maximum likelihood (REML), developed by Patterson and Thompson (1971 ). Priors for the variance components are derived from the VARCOMP-procedure within SAS (SAS, 1985) using REML from a model where the district effect was replaced by a "riding association" effect (3 riding associations; no absorption ). A bivariate mixed model with unequal designs for the traits is used for the REML procedure. Details are described by Meyer (1983). The procedure allows unequal designs for the fixed effects of the two traits, measured on the same animal. Horses with one trait missing can be included. The number of records included in each analysis is shown in Tables 5 and 6. From Table 5 it can be extracted that about 50% have both traits recorded, about 40% have only records for dressage and about 10% have only records for jumping. In The Netherlands most horses have to start in dressage competition before entering TABLE 5 Number of records and horses involved in analyzing the relationship between jumping and dressage with REML Traits
Total number of horses with performances in: Jumping and dressage
Total no. horses
Total no. performances
Only dressage
Only jumping
Progeny of the younger stallions LT4-LT6 1279 LT4-LT7 1682 LT4-LT8 1991
1257 1544 1785
284 349 404
2820 3575 4180
4099 5257 6171
Progeny of the older stallions LT4-LT6 1884 LT4-LT7 2889 LT4-LT8 3748
1519 2217 2884
387 590 756
3790 5696 7388
5674 8585 11136
339 TABLE 6 Number of records and horses involved in analyzing the relationship between performances at 4, 5 and 6 years of age Traits first/second
Total number of horses with performances in: First and second
Only first
Total no. horses
Total no. performances
Only second
Progeny of younger stallions in jumping LT4,LT5 607 385 LT4,LT6 340 652 LT4,LT6 940 535
868 972 372
1860 1964 1847
2467 2304 2787
Progeny of older stallions in jumping LT4,LT5 1171 378 LT4,LT6 698 851 LT5,LT6 2029 864
1722 2106 775
3271 3655 3668
4442 4353 5697
Progeny of younger stallions in dressage LT4,LT5 1411 663 LT4,LT6 806 1268 LT5,LT6 1663 897
1149 1229 372
3223 3303 2932
4634 4109 4595
Progeny of older stallions in dressage LT4,LT5 2320 642 LT4,LT6 1490 1472 LT5,LT6 3421 1228
2329 2780 849
5291 5742 5498
7611 7232 8919
j u m p i n g c o m p e t i t i o n . T h e r e f o r e , 90% f r o m t h e y o u n g c o m p e t i t i o n horses h a v e p e r f o r m a n c e s in dressage. RESULTS AND DISCUSSION T h e ( c o ) v a r i a n c e c o m p o n e n t e s t i m a t i o n in D a t a s e t A gives e s t i m a t e s of h e r i t a b i l i t y for j u m p i n g a n d dressage a n d e s t i m a t e s of p h e n o t y p i c a n d genetic c o r r e l a t i o n s b e t w e e n t h e s e two traits. T h e results of t h e a n a l y s e s w i t h R E M L are s u m m a r i z e d in T a b l e 7. D a t a f r o m t h e y o u n g e r stallions a n d f r o m t h e i r y o u n g e s t o f f s p r i n g (4-6 y e a r s of age) are e x p e c t e d to give t h e m o s t u n b i a s e d e s t i m a t e d of p a r a m e t e r s . T h e m o s t u n b i a s e d h e r i t a b i l i t y e s t i m a t e s are 0.20 for j u m p i n g a n d 0.10 for dressage. E s t i m a t e s of h e r i t a b i l i t y d e t e r m i n e d f r o m t h e d a t a of t h e older stallions are less c o n s i s t e n t a n d a v e r a g e 0.12 for j u m p i n g a n d 0.05 for dressage. D a t a f r o m the older stallions h a v e b e e n subjected to m o r e selection b e c a u s e selection could also be b a s e d on older o f f s p r i n g of t h e stallions. T h i s h a s r e s u l t e d in t h e b e t t e r stallions h a v i n g r e l a t i v e l y m o r e a n d t h e w o r s e stallions h a v i n g fewer or no
340 TABLE 7 Estimates of heritabilities and correlations between jumping and dressage Dataset
Heritabilities Jumping
Correlations Dressage
Genetic
Phenotypic
Offspring of younger stallions LT4-LT6 0.20 LT4-LT7 0.20 LT4-LT8 0.18
0.10 0.09 0.08
- 0.27 - 0.09 -0.06
0.26 0,26 0.25
Offspring of the older stallions LT4-LT6 0.11 LT4-LT7 0.10 LT4-LT8 0.16
0.06 0.06 0.03
0.00 - 0.04 0.10
0.15 0.19 0.22
progeny in competition. This phenomenon is probably responsible for the low variance among sires in the data derived from offspring sired by the older stallions. Only 30% of the registered foals have competition data. Apart from directional selection for attendance at competition, these low percentages are caused by rearing losses and selection for alternative purposes of the riding horses. Alternative purposes are horses for riding schools and recreation, mares as breeding stock and horses for export. Selection of young horses for alternative purpose is based on phenotypes of traits. When these traits have genetic correlations with performances in competition, which may be possible, these selections might have an impact on the remaining variance available for directional selection for attendance at competition. Selection for attendance at competition will be based on correlated traits. Meyer and Thompson (1984) researched the influence of selection on (co)variance estimates when selection was conducted on a correlated trait. They concluded that maximum likelihood will account for selection bias when all information contributing to selection decision is included in the model of analysis. Otherwise maximum likelihood estimates will be biased as well, nevertheless for a large range of genetic parameters constellations still considerably less than the corresponding analysis of variance estimates. The heritability estimates for dressage are lower than those for jumping. Dressage performance measured at competition may be more affected by environmental variation. Moreover, selection for attendance at competition is more directed to dressage characteristics such as conformation and gaits. Hence, downward bias on heritability estimates is expected to be greater for dressage than for jumping. The heritability estimates, derived from the multiple trait REML analyses
341
based on a model with the district effect, are on average 25% lower than the estimates used as priors. The district effect is significant and is responsible for about 50% of the explained variation. The prior estimates are derived from a model where the district effect is replaced by the "riding association" effect (only 3 classes). The main cause of this discrepancy is probably the partial confounding of the stallions and regions in which the stallions service their mares. Differences in performance level between the regions will be included in the sire variance. This results in a overestimation of the heritability. The phenotypic correlation between jumping and dressage is slightly positive, which is in close agreement with Thafvelin et al. (1980). The genetic correlation between jumping and dressage performances appears to be near zero. This agrees with the indirect estimates, based on correlations between estimated breeding values, from Bruns (1981), and Meinardus and Bruns ( 1987 ). They did not adjust for the repeatabilities of the breeding values of the stallions for these two traits. But, even doing so, Taylor (1982) is stating that this approximation is unstable, often yielding estimates of genetic correlation outside the parameter space. Although the REML estimates are not different from the former indications, they are preferred. Meyer and Thompson (1984) simulated data with a large range of genetic parameter constellations where culling was on a correlated trait. They found slightly reduced genetic correlations when analysis was with maximum likelihood. Although parameter constellation of our datasets is unknown, it can be expected that selection will have only minor impact on the estimates of genetic correlation between jumping and dressage. Further, stallions often have different popularities. Preferential treatment of the offspring of stallions for one or the other trait will bias the genetic correlation downwards. The inclusion of the district effect (inclusive the riding clubs ) in our model will partially adjust for this bias. It is not possible to quantify the exact impact of preferential treatment and to qualify whether or not it is different for the offspring from the younger and older stallions. Data from central performance tests of stallions can probably give less biased estimates for correlations between traits. Bruns et al. (1985) estimated correlations under those conditions and their range of estimates covers the estimates from our data. Estimates of genetic correlation between jumping and dressage slightly increase when data are expanded by adding the 7- and 8-year-old offspring. This tendency might be explained by the growing or declining popularity of the stallions with age of the progeny group. This could influence the relative quality of the riders of a progeny group. The (co)variance component estimation in Dataset B gives estimates of phenotypic and genetic correlations between the separate lifetime totals (LTs) at different ages (Table 8). The genetic correlations between performances at 4, 5 and 6 years are not different for the data sets. The average genetic corre-
342 TABLE 8 Estimates correlations (p,g) between performances at 4, 5 and 6 year of age Younger stallions LT4 ,Jumping LT4 LT5 LT6
0.95 0.89
Dressage LT4 LT5 LT6
0.90 0.96
Older stallions
LT5
LT6
0.68
0.65 0.85
1.05 0.78 1.03
0.69 0.90
LT4
0.85 0.93
LT5
LT6
0.64
0.55 0.81
0.96 0.78
0.80 1.08
0.65 0.87
0.99
L T , = Lifetime total at an age of i years; g = genetic correlations lower triangle; p = p h e n o t y p i c correlations upper triangle.
lation between LT4 and LT5 is 0.88 and between LT4 and LT6 is 0.97. Between LT5 and LT6 it averages 1.01. These estimates are slightly higher than those given by Tavernier (1986). The trait lifetime total (LT) is, however, cumulatively built up. Hence, autocorrelation is involved. Nevertheless, it can be concluded that the level of the cumulative performances at 4, 5 and 6 years of age are controlled by almost the same genes. Phenotypic correlations are also consistent for these data sets. The average phenotypic correlation between LT4 and LT5 is 0.72 and between LT4 and LT6 is 0.67, while the correlation between LT5 and LT6 amounts to 0.86. These estimates are higher than the repeatabilities given in Table 1, which is probably also a result of the autocorrelation between lifetime totals. The estimates of heritability of the separate lifetime totals, which are not shown, are similar to those of the combined performance traits in Table 7. The heritability estimates of data from the younger stallions are slightly lower than those in Table 7. This can be ascribed to a selection in the linked data which consists of records from previous years (1983, 1984, 1985). CONCLUSIONS
(1) This study suggests that estimates derived from data based on the youngest offspring from the younger stallions are most desirable. They are the least affected by selection. Further research should establish the expected selection bias for analysis with various selection and parameter constellations. Approaches to adjust (partially) for selection should be explored. (2) Competition data on jumping give moderately high (0.20) heritability
343
estimates. Competition data on dressage give low heritability estimates (<0.10). (3) Genetic evaluation without including an important fixed effect such as districts, or riding clubs, will be biased owing to differences between these locations or clubs. (4) Jumping and dressage have probably little phenotypic or any genetic relationship. (5) The high genetic correlation between performances at different ages implies that genetic evaluation of breeding candidates based on performances at an early age will be effective in making genetic progress in performance at all ages. The generation interval will be as short as possible. Early proofs for stallions for example will be less affected by non-random mating, selection and preferential treatment of the offspring. ACKNOWLEDGEMENTS
The present study was made possible by "Stichting Fonds Nederlandse Veefokkerij" and the Royal Warmblood Studbook of the Netherlands. We are grateful to Dr. S. Korver for helpful comments and to Dr. K. Meyer for providing the computer program.
REFERENCES Bruns, E., 1981. Estimation of the breeding value of stallions from the tournament performance of their offspring. Livest. Prod. Sci., 8: 465-473. Bruns, E., Rauls, B. and Bade, B., 1985. Die Entwicklung von Selektionskriterien fur die Reitpferdezucht. V. Phanotypische und genetische Parameter und Selektionsindices fur eigenleistungsgeprufte Hengste. Zuchtungskunde, 57: 172-182. Huizinga, H.A., 1987. Genetische analyse van spring- en dressuurprestaties in de warmblood paardenpopulatie in Nederland. Vakgroep ZoStechniek, Rijksuniversiteit van Utrecht. Faculteit Diergeneeskunde, 66 pp. In de Strengen, 1987. WPN-hengstenstapel 1987.54-ste jaargang, no. 6, pp. 70-84. Langlois, B., 1980. Estimation de la valeur de g4n~tique des chevaux de sport d'apr~s les sommes gagn4es dans les comp4titions 6questres francaises. Ann. Genet. Sel. Anim., 12: 15-31. Meinardus, H. and Bruns, E., 1987. BLUP-procedure in riding horses based on competition results. EAAP, Lisbon, Portugal, 28 September-1 October. Meyer, K., 1983. Maximum likelihood procedures for estimating genetic parameters for later lactations of dairy cattle. J. Dairy Sci., 66: 1988-1997. Meyer, K. and Thompson, R., 1984. Bias in variance and covariance estimators due to selection on a correlated trait. Z. Tierz. Zuechtungsbiol., 101: 33-50. Patterson, H.D. and Thompson, R., 1971. Recovery of inter-block information when block sizes are unequal. Biometrika, 58: 545-554. Philipsson, J., 1975. Estimates of heritability for performances of Swedish riding horses. E.A.A.P., Warsaw, Poland, 23-27 June. SAS, 1985. User's Guide! Statistics, 5. Cary, NC, 956 pp.
344 Tavernier, A., 1986. Donn~es nouvelles sur les performances des cheveaux de sport: precocitY, effects maternels, influence du type g~n~tique. 12~ Journ~e d'~tude, pp. 31-57. Tavernier, A., 1986. Estimation of breeding value with main emphasis on stock mare candidates. E.A.A.P., Budapest, Hungary, 1-4 September 1986. Taylor, J.F., 1982. Assumptions required to approximate unbiased estimates of genetic (co)variance by the method of Calo et al. (1973). Genet. Res., 1982-1983; 256-261. Thafvelin, B., Philipsson, J. and Darenius, A., 1980. Genetic study on riding horse traits under field conditions. E.A.A.P., Munchen, 1-4 September.
RESUME Huizinga, H.A. et van der Meij, G.J.W., 1989. Estimation des param~tres des performances en saut et en competition de dressage des chevaux de selle N~erlandais. Livest. Prod. Sci., 21: 333345 (en anglais). Cette ~tude a pour objectif d'estimer plusieurs param~tres g~n6tiques de la population des chevaux de selle N~erlandais. Les caract~res concern~s sont les performances en saut d'obstacles et en dressage et les param~tres estim~s sont: l'h~ritabilit~ pour le saut et le dressage, les correlations ph~notypiques et g6n~tiques entre le saut et le dressage et les correlations ph~notypiques et g~ndtiques entre les performances h diff~rents figes. On a utilis$ la m~thode du maximum de vraisemblance restreint (REML). Les donnges concernent 6899 chevaux pour le saut et 10 408 pour le dressage. Ces chevaux sont issus de 205 et 237 ~talons respectivement et leur ~ge varie de 4 h 8 ans. Le caract~re de performance est une note cumul~e qui traduit les r4sultats en competition. La racine carrie de cette note est la plus appropri4e pour normaliser les donn~es. Pour l'estimation des param~tres ph~notypiques et g~n~tiques les donn~es ont dt~ r@arties en deux groupes selon l'~ge des p~res. Pour l'estimation des correlations entre les performances h 4,5 et h 6 ans, les performances des ann~es ant~rieures ont ~t~ jointes aux donn~es. Les estimations les moins biais~es de l'h~ritabilit~ pour le saut et le dressage sont fournies par les chevaux les plus jeunes issus des ~talons les plus jeunes; elles sont de 0,210et 0,10 respectivement. Les correlations g~n~tiques entre le saut et le dressage varient de - 0,2~ h 0,10. Les correlations ph6notypiques correspondantes se situent entre 0,15 et 0,26. Les corr~ations ph~notypiques et g~n~tiques entre les performances h 4,5 et 6 ans s'~l~vent en moyenne h 0,9'5 et 0,75 respectivement. Ces derni~res valeurs sont d'un grand int~rSt pour l'~valuation gdn~tique!des candidats h la reproduction dans la population. KURZFASSUNG Huizinga, H.A. and van der Meij, G.J.W., 1989. Schiitzung von Parametern der Leistung holl~indischer Warmblutpferde in Spring- und Dressur-Wettbewerben. Livest. Prod. Sci. 21:333-345 (auf englisch). Ziel der Arbeit ist die Sch~itzung verschiedener genetischer Parameter in der holl~indischen Warmblut-Reitpferde°Population.Die berticksichtigten Merkmale sind die Leistungen in Springund Dressur-Wettbewerben. Folgende Parameter wurden geschiitzt: Heritabilit~it for die Spring- und Dressurleistung und genetische Korrelationen zwischen diesen, phiinotypische und genetische Korrelationen zwischen den Leistungen bei verschiedenem Alter. Ftir die Parameter-Schiitzung wurde die Methode "Restricted Maximum Likelihood (REML)" angewandt.
345 Es standen Daten von 6899 Pferden mit Springleistungen und 10 408 Pferden mit Dressurergebnissen zur Verfiigung, die von 205 bzw. 237 Hengsten abstammen. Das Alter der Nachkommen variierte zwischen 4 und 8 Jahren. Als Leistungsergebnis wurde eine kumulativ abgeleitete Note verwendet, die das Niveau der Leistung im Wettbewerb widerspiegelt. Eine Quadratwurzel-Transformation wurde verwendet, um eine Normalverteilung der Daten zu erreichen. Fiir die Schiitzung der phiinotypischen und genetischen Parameter wurden die Daten in zwei Datens~itze entsprechend des Alters der V~iter aufgeteilt. Ftir die Beziehung zwischen Leistungen bei 4, 5 und 6 Jahren wurden die Leistungen der Nachkommen friiherer Jahre mit den vorliegenden Daten verkniipft. Die unverzerrtesten Heritabilit~itsschiitzwerteflit die Spring- und Dressurleistung ergaben sich ftir die jiingsten Nachkommen der jiingsten Hengste mit 0,20 bzw. 0,10. Die genetische Korrelation zwischen der Spring- und Dressurleistung lag im Bereich v o n - 0,27 und + 0,10. Die ph~notypische Korrelation zwischen diesen Merkmalen schwankte zwischen + 0,15 und + 0,26. Phiinotypische und genetische Korrelationen zwischen den Leistungen bei 4, 5 und 6 Jahren lagen im Mittel bei 0,95 bwz. 0,75. Die zuletzt aufgefiihrten Ergebnisse haben bedeutende Auswirkungen auf die Zuchtwertschiitzung in der Population.
333
Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands
Estimated Parameters of Performance in Jumping and Dressage Competition of the Dutch Warmblood Horse H.A. HUIZINGA 1 and G.J.W. VAN DER MEIJ 2
1Department of Animal Breeding, Wageningen Agricultural University, P.O. Box 338, 6700 AH Wageningen (The Netherlands) 2Department of Herd Health and Reproduction, Faculty o[ Veterinary Medicine, University of Utrecht, P.O. Box 80.151, Utrecht (The Netherlands) (Accepted for publication 28 November 1988)
ABSTRACT Huizinga, H.A. and van der Meij, G.J.W., 1989. Estimated parameters of performance in jumping and dressage competition of the Dutch Warmblood horse. Livest. Prod. Sci., 21: 333-345. The objective of this study is to estimate several genetic parameters in the Dutch Warmblood riding horse population. The traits involved are performances in jumping and dressage competition. The following parameters are estimated: heritabilities for jumping and dressage; phenotypic and genetic correlations between jumping and dressage; and phenotypic and genetic correlations between performances at different ages. These parameters are estimated by restricted maximum likelihood (REML). Data are from 6899 horses with performances in jumping and 10 408 horses with performances in dressage competition. The horses are sired by 205 and 237 stallions for the two traits, respectively. The progeny range in age from 4 to 8 years old. The performance trait is a cumulatively derived score, that reflects the level of performance in competition. A square root transformation of the score is most appropriate to normalize the data. For estimation of phenotypic and genetic parameters the data is split into two data sets according to the age of the sires {offspring sired by older vs. younger stallions). For estimating correlations between performances at 4, 5 and 6 years of age, performances of the offspring out of previous years are linked to the data. The most unbiased estimates of heritability for jumping and dressage are from data derived from the youngest offspring sired by the younger stallions and are 0.20 and 0.10, respectively. Genetic correlation between jumping and dressage ranges from -0.27 to 0.10. The phenotypic correlation between these traits ranges from 0.15 to 0.26. Phenotypic and genetic correlations between performances at 4, 5 and 6 years average 0.95 and 0.75, respectively. These latter results have important implications for genetic evaluation of breeding candidates in the population.
INTRODUCTION The active breeding population of the Dutch riding horse consists of about 9 0 0 0 m a t e d m a r e s y e a r - 1 w i t h s e r v i c e s f r o m a b o u t 150 s t a l l i o n s ( I n t h e S t r e n -
0301-6226/89/$03.50
© 1989 Elsevier Science Publishers B.V.
334 gen, 1987). Horses are registered by the Royal Warmblood Studbook of the Netherlands (KWPN). To predict the response to selection, to evaluate various breeding plans and to predict breeding values of breeding candidates for selection, a knowledge of genetic parameters is essential. The first phase in developing a breeding program is defining the breeding goal. Even when evaluation is restricted to performance traits, we still have to deal with a least two traits; jumping and dressage. Estimates of heritability of these traits in the Dutch Warmblood horse population are needed. Breeders and riders are frequently striving for simultaneous improvement in both traits. Before considering the combination of jumping and dressage in the breeding goal, knowledge of phenotypic and genetic (co)variances is of interest. Progeny testing in general will increase generation interval. To limit this disadvantage, e.g. stallions can be evaluated on the basis of performances of their first offspring at an early age. For justifying this, knowledge is needed of genetic correlations between performances at younger and older ages of the horse. Summarizing, there are three objectives in this paper: (1) to estimate heritability for jumping and dressage; (2) to estimate relationships between jumping and dressage; (3) to estimate relationships between the level of performance at different ages. REVIEW OF LITERATURE In France, earnings per year (Langlois, 1980; Tavernier, 1986) and in the F.R.G. earnings per start (Bruns, 1981; Meinardus and Bruns, 1987) in competition are the traits that are involved in the genetic evaluation of breeding candidates in the population. In Sweden, the number of placings and the number of starts are the criteria (Philipsson, 1975). Most procedures adjust for age, sex and year of performance. Riding club effects, which are expected in The Netherlands, were not taken into account in these studies. A review of estimates of heritability for jumping and dressage is given in Table 1. Estimates of heritability are in most cases higher for jumping than for dressage. However, Bruns (1981) found the opposite. Estimates of phenotypic correlation under field conditions assessed by Thafvelin et al. (1980) between gaits under saddle and jumping ability ranged from 0.23 to 0.30. Phenotypic correlations, which can be considered as repeatabilities, are reported by a number of studies, as shown in Table 1. Bruns (1981) and Meinardus and Bruns (1987) estimated genetic correlations between jumping and dressage indirectly by computing correlations between breeding values for the two traits. Estimates were up to 0.10 and 0.14, respectively. Bruns et al. ( 1985 ) estimated parameters from data from central performance tests of stallions. The estimates for genetic correlation between
335 TABLE 1 Estimates of heritabilities and repeatabilities for jumping and dressage in France, F.R.G. and Sweden Heritabilities
Langlois (1980) Tavernier (1986) Bruns (1981) Meinardus and Bruns (1987) Philipsson (1975)
Repeatabilities
Jumping
Dressage
Jumping
Dressage
0.15-0.19 0.20-0.25 0.13-0.14 0.18-0.21 0.16-0.29
-0.12-0.05
0.46-0.53 0.45 0.21-0.34
0.46-0.54 0.29-0.39
0.42-0.48
0.45-0.48
0.15-0.27 0.16-0.18 0.04-0.06
show jumping and gaits averaged 0.37. Their estimates for genetic correlation between free jumping and gaits averaged - 0.53 and between free jumping and riding ability - 0.06. Tavernier (1986) estimated genetic correlations between performances in jumping competition from 4 through 10 years of age. The genetic correlations averaged 0.75 between 4 and the other ages, 0.85 between 5 and the other ages and 1 between the remaining age classes. MATERIALS AND METHODS
For estimation of phenotypic and genetic parameters two datasets, A and B, are created. Dataset A is based on competition results from 1986. In these data, heritabilities for jumping and dressage, and phenotypic and genetic correlations between jumping and dressage are estimated. Dataset B is based on horses with competition results in 1986, while repeated observations from previous years are linked to this dataset. These repeated measures of the performances at an earlier age enhance the number of records considerably, while the number of horses involved did not change. These data are used for estimating phenotypic and genetic correlations between performances at different ages of the horse. Records from offspring over 8 years old in 1986 and data of stallions with less than 5 offspring are excluded. The remaining data sets contain 6899 horses in jumping competition sired by 204 stallions and 10 408 horses in dressage competition sired by 237 stallions. Table 2 shows the number of records year- 1 in the data sets. The trait which expresses performance in sport competition is the so-called "highest level" in sport, accomplished during the lifetime of the horse. The lifetime total is cumulative and reflects the level of performance of the horse
336 TABLE 2 Number of records per year in the data Number ofhorses in 1986
Jumping Dressage Content of Datasets: A
6899 10408
B
X
Number of repeated measures in: 1985
1984
1983
4696 7641
2628 4608
1107 2228
X
X
X
X
TABLE 3 Means and standard deviations for the traits in the two datasets Jumping
Dressage
Younger
Older
Younger
Older
Dataset A LT4-LT6 LT4-LT7 LT4-LT8
2.57 _+1.05 2.61 +- 1.07 2.67 +_1.09
2.65 _+1.05 2.74+_ 1.07 2.80 +_1.10
2.21 +- 1.16 2.26+_ 1.17 2.30 +_1.18
2.18 +- 1.16 2.28+_ 1.18 2.35 +_1.19
Dataset B LT4 LT5 LT6 LT7 LT8
2.17 +_0.93 2.54 +_1.03 2.75 +- 1.09 2.86+_ 1.12 2.99 +- 1.14
2.19 +-0.92 2.59 +_1.01 2.82 +- 1.10 2.95 +_1.11 3.02 +- 1.15
1.79 +_1.12 2.17 +_1.14 2.38+_ 1.17 2.47 +_1.19 2.58 +- 1.21
1.78 +- 1.12 2.13 +_1.15 2.36+_ 1.16 2.50_+ 1.19 2.58 +_1.23
in a discipline of competition. These scores, consisting of a character and number combination, are transformed to a linear scale. For a more detailed descript i o n o f t h e t r a n s f o r m a t i o n o f s c o r e s , see H u i z i n g a ( 1 9 8 7 ) . T o a c c o m p l i s h a m o r e n o r m a l l y d i s t r i b u t e d r e s i d u a l e r r o r t e r m , a s q u a r e r o o t t r a n s f o r m a t i o n is applied. I t is e x p e c t e d t h a t d a t a f r o m t h e o l d e r s t a l l i o n s h a v e b e e n s u b j e c t e d t o m o r e selection. Therefore, data are split into two groups according to birth date of t h e s t a l l i o n ( b e f o r e 1972 a n d 1972 o r a f t e r ) . M e a n s a n d s t a n d a r d d e v i a t i o n s f o r t h e t r a i t s i n t h e d i f f e r e n t d a t a s e t s a r e g i v e n i n T a b l e 3. From 4 to 5 years of age about 85% of the horses make progress during a c o m p e t i t i o n y e a r . F r o m 7 t o 8 y e a r s o f a g e o n l y a b o u t 35% a r e a b l e t o r e a l i z e p r o g r e s s . W h e n a h i g h l e v e l o f p e r f o r m a n c e is r e a c h e d , m a k i n g p r o g r e s s b e comes more difficult. But also factors related to breeding, export and culling
337 TABLE 4 Number of records and sires per dataset per trait Jumping
Dressage
Younger
Older
Younger
Older
DatasetA LT4-LT6 LT4-LT7 LT4-LT8
1563/ 94 2031/100 2396/101
2271/ 91 3479/100 4504/104
2536/103 3226/111 3778/114
3403/ 99 5106/115 6632/123
DatasetB LT4 LT5 LT6 LT7 LT8
992/ 1472/ 1312/ 800/ 364/
1549/ 97 2893/103 2804/100 2132/ 96 1025/ 90
2074/110 2560/105 2035/ 88 1206/ 69 550/ 60
2962/110 4649/122 4270/119 3092/115 1526/106
95 95 90 64 52
play a role, and for these reasons the datasets are split according to the age of the offspring. Only for Dataset A is this of importance and leads to three groups of records referring to lifetime totals (LT's) based on age: lifetime totals from 4 through 6 years; lifetime totals from 4 through 7 years; lifetime totals from 4 through 8 years. Dataset B is analysed within age classes. The respective lifetime totals are at 4 ( LT4 ), 5 (LT5) and 6 ( LT6 ) years of age. A description of the data results in frequencies as shown in Table 4. The following linear models are applied for analyses of the performance records of jumping and dressage. For Dataset A: Y~jt~,, = district~ + agei + sexl + sire~ + e,~l~o
For Dataset B: Y;kt.... = district~ + NYSk + sext + yearm + sire~ + eikl~no where: = observed score on the oth offspring; Y~jtno,, Yim . . . . District,. = effect of the ith district ( i = 1 .... 200); Agej = e f f e c t o f t h e j t h age-class (1"=4, 5, 6,); NYS~ ---effect of the number of years in competition (k--1, 2, 3); ---effect of the year ( m = 1983, 1984, 1985, 1986); Yearm Sex/ = effect of t h e / t h sex (l = stallions, mares, geldings); = random effect of the nth sire; Siren - r a n d o m residual term. e ijlno'e iklmno In the combined data of the younger and older stallions, 780 riding clubs for
338
jumping and 920 riding clubs for dressage are represented. For jumping as well as for dressage 62.8% of the riding clubs have less then 11 combinations participating in competition. By dividing the data according to age of the sires, the number of observations per riding club is further decreased. To increase these numbers, the number of riding clubs is reduced to 200 districts. Geographical and data similarities are used to join the riding clubs. Statistical methods generally used in the analysis of animal breeding assume that these data are a random sample from the population. However, this is often not true, also not for competition data from horses. From 137 traced stallions, covering 75% of the data with their offspring, only about 30% of the registered foals are brought into dressage competition and 20% into the jumping competition. Variance and covariance components are estimated by restricted maximum likelihood (REML), developed by Patterson and Thompson (1971 ). Priors for the variance components are derived from the VARCOMP-procedure within SAS (SAS, 1985) using REML from a model where the district effect was replaced by a "riding association" effect (3 riding associations; no absorption ). A bivariate mixed model with unequal designs for the traits is used for the REML procedure. Details are described by Meyer (1983). The procedure allows unequal designs for the fixed effects of the two traits, measured on the same animal. Horses with one trait missing can be included. The number of records included in each analysis is shown in Tables 5 and 6. From Table 5 it can be extracted that about 50% have both traits recorded, about 40% have only records for dressage and about 10% have only records for jumping. In The Netherlands most horses have to start in dressage competition before entering TABLE 5 Number of records and horses involved in analyzing the relationship between jumping and dressage with REML Traits
Total number of horses with performances in: Jumping and dressage
Total no. horses
Total no. performances
Only dressage
Only jumping
Progeny of the younger stallions LT4-LT6 1279 LT4-LT7 1682 LT4-LT8 1991
1257 1544 1785
284 349 404
2820 3575 4180
4099 5257 6171
Progeny of the older stallions LT4-LT6 1884 LT4-LT7 2889 LT4-LT8 3748
1519 2217 2884
387 590 756
3790 5696 7388
5674 8585 11136
339 TABLE 6 Number of records and horses involved in analyzing the relationship between performances at 4, 5 and 6 years of age Traits first/second
Total number of horses with performances in: First and second
Only first
Total no. horses
Total no. performances
Only second
Progeny of younger stallions in jumping LT4,LT5 607 385 LT4,LT6 340 652 LT4,LT6 940 535
868 972 372
1860 1964 1847
2467 2304 2787
Progeny of older stallions in jumping LT4,LT5 1171 378 LT4,LT6 698 851 LT5,LT6 2029 864
1722 2106 775
3271 3655 3668
4442 4353 5697
Progeny of younger stallions in dressage LT4,LT5 1411 663 LT4,LT6 806 1268 LT5,LT6 1663 897
1149 1229 372
3223 3303 2932
4634 4109 4595
Progeny of older stallions in dressage LT4,LT5 2320 642 LT4,LT6 1490 1472 LT5,LT6 3421 1228
2329 2780 849
5291 5742 5498
7611 7232 8919
j u m p i n g c o m p e t i t i o n . T h e r e f o r e , 90% f r o m t h e y o u n g c o m p e t i t i o n horses h a v e p e r f o r m a n c e s in dressage. RESULTS AND DISCUSSION T h e ( c o ) v a r i a n c e c o m p o n e n t e s t i m a t i o n in D a t a s e t A gives e s t i m a t e s of h e r i t a b i l i t y for j u m p i n g a n d dressage a n d e s t i m a t e s of p h e n o t y p i c a n d genetic c o r r e l a t i o n s b e t w e e n t h e s e two traits. T h e results of t h e a n a l y s e s w i t h R E M L are s u m m a r i z e d in T a b l e 7. D a t a f r o m t h e y o u n g e r stallions a n d f r o m t h e i r y o u n g e s t o f f s p r i n g (4-6 y e a r s of age) are e x p e c t e d to give t h e m o s t u n b i a s e d e s t i m a t e d of p a r a m e t e r s . T h e m o s t u n b i a s e d h e r i t a b i l i t y e s t i m a t e s are 0.20 for j u m p i n g a n d 0.10 for dressage. E s t i m a t e s of h e r i t a b i l i t y d e t e r m i n e d f r o m t h e d a t a of t h e older stallions are less c o n s i s t e n t a n d a v e r a g e 0.12 for j u m p i n g a n d 0.05 for dressage. D a t a f r o m the older stallions h a v e b e e n subjected to m o r e selection b e c a u s e selection could also be b a s e d on older o f f s p r i n g of t h e stallions. T h i s h a s r e s u l t e d in t h e b e t t e r stallions h a v i n g r e l a t i v e l y m o r e a n d t h e w o r s e stallions h a v i n g fewer or no
340 TABLE 7 Estimates of heritabilities and correlations between jumping and dressage Dataset
Heritabilities Jumping
Correlations Dressage
Genetic
Phenotypic
Offspring of younger stallions LT4-LT6 0.20 LT4-LT7 0.20 LT4-LT8 0.18
0.10 0.09 0.08
- 0.27 - 0.09 -0.06
0.26 0,26 0.25
Offspring of the older stallions LT4-LT6 0.11 LT4-LT7 0.10 LT4-LT8 0.16
0.06 0.06 0.03
0.00 - 0.04 0.10
0.15 0.19 0.22
progeny in competition. This phenomenon is probably responsible for the low variance among sires in the data derived from offspring sired by the older stallions. Only 30% of the registered foals have competition data. Apart from directional selection for attendance at competition, these low percentages are caused by rearing losses and selection for alternative purposes of the riding horses. Alternative purposes are horses for riding schools and recreation, mares as breeding stock and horses for export. Selection of young horses for alternative purpose is based on phenotypes of traits. When these traits have genetic correlations with performances in competition, which may be possible, these selections might have an impact on the remaining variance available for directional selection for attendance at competition. Selection for attendance at competition will be based on correlated traits. Meyer and Thompson (1984) researched the influence of selection on (co)variance estimates when selection was conducted on a correlated trait. They concluded that maximum likelihood will account for selection bias when all information contributing to selection decision is included in the model of analysis. Otherwise maximum likelihood estimates will be biased as well, nevertheless for a large range of genetic parameters constellations still considerably less than the corresponding analysis of variance estimates. The heritability estimates for dressage are lower than those for jumping. Dressage performance measured at competition may be more affected by environmental variation. Moreover, selection for attendance at competition is more directed to dressage characteristics such as conformation and gaits. Hence, downward bias on heritability estimates is expected to be greater for dressage than for jumping. The heritability estimates, derived from the multiple trait REML analyses
341
based on a model with the district effect, are on average 25% lower than the estimates used as priors. The district effect is significant and is responsible for about 50% of the explained variation. The prior estimates are derived from a model where the district effect is replaced by the "riding association" effect (only 3 classes). The main cause of this discrepancy is probably the partial confounding of the stallions and regions in which the stallions service their mares. Differences in performance level between the regions will be included in the sire variance. This results in a overestimation of the heritability. The phenotypic correlation between jumping and dressage is slightly positive, which is in close agreement with Thafvelin et al. (1980). The genetic correlation between jumping and dressage performances appears to be near zero. This agrees with the indirect estimates, based on correlations between estimated breeding values, from Bruns (1981), and Meinardus and Bruns ( 1987 ). They did not adjust for the repeatabilities of the breeding values of the stallions for these two traits. But, even doing so, Taylor (1982) is stating that this approximation is unstable, often yielding estimates of genetic correlation outside the parameter space. Although the REML estimates are not different from the former indications, they are preferred. Meyer and Thompson (1984) simulated data with a large range of genetic parameter constellations where culling was on a correlated trait. They found slightly reduced genetic correlations when analysis was with maximum likelihood. Although parameter constellation of our datasets is unknown, it can be expected that selection will have only minor impact on the estimates of genetic correlation between jumping and dressage. Further, stallions often have different popularities. Preferential treatment of the offspring of stallions for one or the other trait will bias the genetic correlation downwards. The inclusion of the district effect (inclusive the riding clubs ) in our model will partially adjust for this bias. It is not possible to quantify the exact impact of preferential treatment and to qualify whether or not it is different for the offspring from the younger and older stallions. Data from central performance tests of stallions can probably give less biased estimates for correlations between traits. Bruns et al. (1985) estimated correlations under those conditions and their range of estimates covers the estimates from our data. Estimates of genetic correlation between jumping and dressage slightly increase when data are expanded by adding the 7- and 8-year-old offspring. This tendency might be explained by the growing or declining popularity of the stallions with age of the progeny group. This could influence the relative quality of the riders of a progeny group. The (co)variance component estimation in Dataset B gives estimates of phenotypic and genetic correlations between the separate lifetime totals (LTs) at different ages (Table 8). The genetic correlations between performances at 4, 5 and 6 years are not different for the data sets. The average genetic corre-
342 TABLE 8 Estimates correlations (p,g) between performances at 4, 5 and 6 year of age Younger stallions LT4 ,Jumping LT4 LT5 LT6
0.95 0.89
Dressage LT4 LT5 LT6
0.90 0.96
Older stallions
LT5
LT6
0.68
0.65 0.85
1.05 0.78 1.03
0.69 0.90
LT4
0.85 0.93
LT5
LT6
0.64
0.55 0.81
0.96 0.78
0.80 1.08
0.65 0.87
0.99
L T , = Lifetime total at an age of i years; g = genetic correlations lower triangle; p = p h e n o t y p i c correlations upper triangle.
lation between LT4 and LT5 is 0.88 and between LT4 and LT6 is 0.97. Between LT5 and LT6 it averages 1.01. These estimates are slightly higher than those given by Tavernier (1986). The trait lifetime total (LT) is, however, cumulatively built up. Hence, autocorrelation is involved. Nevertheless, it can be concluded that the level of the cumulative performances at 4, 5 and 6 years of age are controlled by almost the same genes. Phenotypic correlations are also consistent for these data sets. The average phenotypic correlation between LT4 and LT5 is 0.72 and between LT4 and LT6 is 0.67, while the correlation between LT5 and LT6 amounts to 0.86. These estimates are higher than the repeatabilities given in Table 1, which is probably also a result of the autocorrelation between lifetime totals. The estimates of heritability of the separate lifetime totals, which are not shown, are similar to those of the combined performance traits in Table 7. The heritability estimates of data from the younger stallions are slightly lower than those in Table 7. This can be ascribed to a selection in the linked data which consists of records from previous years (1983, 1984, 1985). CONCLUSIONS
(1) This study suggests that estimates derived from data based on the youngest offspring from the younger stallions are most desirable. They are the least affected by selection. Further research should establish the expected selection bias for analysis with various selection and parameter constellations. Approaches to adjust (partially) for selection should be explored. (2) Competition data on jumping give moderately high (0.20) heritability
343
estimates. Competition data on dressage give low heritability estimates (<0.10). (3) Genetic evaluation without including an important fixed effect such as districts, or riding clubs, will be biased owing to differences between these locations or clubs. (4) Jumping and dressage have probably little phenotypic or any genetic relationship. (5) The high genetic correlation between performances at different ages implies that genetic evaluation of breeding candidates based on performances at an early age will be effective in making genetic progress in performance at all ages. The generation interval will be as short as possible. Early proofs for stallions for example will be less affected by non-random mating, selection and preferential treatment of the offspring. ACKNOWLEDGEMENTS
The present study was made possible by "Stichting Fonds Nederlandse Veefokkerij" and the Royal Warmblood Studbook of the Netherlands. We are grateful to Dr. S. Korver for helpful comments and to Dr. K. Meyer for providing the computer program.
REFERENCES Bruns, E., 1981. Estimation of the breeding value of stallions from the tournament performance of their offspring. Livest. Prod. Sci., 8: 465-473. Bruns, E., Rauls, B. and Bade, B., 1985. Die Entwicklung von Selektionskriterien fur die Reitpferdezucht. V. Phanotypische und genetische Parameter und Selektionsindices fur eigenleistungsgeprufte Hengste. Zuchtungskunde, 57: 172-182. Huizinga, H.A., 1987. Genetische analyse van spring- en dressuurprestaties in de warmblood paardenpopulatie in Nederland. Vakgroep ZoStechniek, Rijksuniversiteit van Utrecht. Faculteit Diergeneeskunde, 66 pp. In de Strengen, 1987. WPN-hengstenstapel 1987.54-ste jaargang, no. 6, pp. 70-84. Langlois, B., 1980. Estimation de la valeur de g4n~tique des chevaux de sport d'apr~s les sommes gagn4es dans les comp4titions 6questres francaises. Ann. Genet. Sel. Anim., 12: 15-31. Meinardus, H. and Bruns, E., 1987. BLUP-procedure in riding horses based on competition results. EAAP, Lisbon, Portugal, 28 September-1 October. Meyer, K., 1983. Maximum likelihood procedures for estimating genetic parameters for later lactations of dairy cattle. J. Dairy Sci., 66: 1988-1997. Meyer, K. and Thompson, R., 1984. Bias in variance and covariance estimators due to selection on a correlated trait. Z. Tierz. Zuechtungsbiol., 101: 33-50. Patterson, H.D. and Thompson, R., 1971. Recovery of inter-block information when block sizes are unequal. Biometrika, 58: 545-554. Philipsson, J., 1975. Estimates of heritability for performances of Swedish riding horses. E.A.A.P., Warsaw, Poland, 23-27 June. SAS, 1985. User's Guide! Statistics, 5. Cary, NC, 956 pp.
344 Tavernier, A., 1986. Donn~es nouvelles sur les performances des cheveaux de sport: precocitY, effects maternels, influence du type g~n~tique. 12~ Journ~e d'~tude, pp. 31-57. Tavernier, A., 1986. Estimation of breeding value with main emphasis on stock mare candidates. E.A.A.P., Budapest, Hungary, 1-4 September 1986. Taylor, J.F., 1982. Assumptions required to approximate unbiased estimates of genetic (co)variance by the method of Calo et al. (1973). Genet. Res., 1982-1983; 256-261. Thafvelin, B., Philipsson, J. and Darenius, A., 1980. Genetic study on riding horse traits under field conditions. E.A.A.P., Munchen, 1-4 September.
RESUME Huizinga, H.A. et van der Meij, G.J.W., 1989. Estimation des param~tres des performances en saut et en competition de dressage des chevaux de selle N~erlandais. Livest. Prod. Sci., 21: 333345 (en anglais). Cette ~tude a pour objectif d'estimer plusieurs param~tres g~n6tiques de la population des chevaux de selle N~erlandais. Les caract~res concern~s sont les performances en saut d'obstacles et en dressage et les param~tres estim~s sont: l'h~ritabilit~ pour le saut et le dressage, les correlations ph~notypiques et g6n~tiques entre le saut et le dressage et les correlations ph~notypiques et g~ndtiques entre les performances h diff~rents figes. On a utilis$ la m~thode du maximum de vraisemblance restreint (REML). Les donnges concernent 6899 chevaux pour le saut et 10 408 pour le dressage. Ces chevaux sont issus de 205 et 237 ~talons respectivement et leur ~ge varie de 4 h 8 ans. Le caract~re de performance est une note cumul~e qui traduit les r4sultats en competition. La racine carrie de cette note est la plus appropri4e pour normaliser les donn~es. Pour l'estimation des param~tres ph~notypiques et g~n~tiques les donn~es ont dt~ r@arties en deux groupes selon l'~ge des p~res. Pour l'estimation des correlations entre les performances h 4,5 et h 6 ans, les performances des ann~es ant~rieures ont ~t~ jointes aux donn~es. Les estimations les moins biais~es de l'h~ritabilit~ pour le saut et le dressage sont fournies par les chevaux les plus jeunes issus des ~talons les plus jeunes; elles sont de 0,210et 0,10 respectivement. Les correlations g~n~tiques entre le saut et le dressage varient de - 0,2~ h 0,10. Les correlations ph6notypiques correspondantes se situent entre 0,15 et 0,26. Les corr~ations ph~notypiques et g~n~tiques entre les performances h 4,5 et 6 ans s'~l~vent en moyenne h 0,9'5 et 0,75 respectivement. Ces derni~res valeurs sont d'un grand int~rSt pour l'~valuation gdn~tique!des candidats h la reproduction dans la population. KURZFASSUNG Huizinga, H.A. and van der Meij, G.J.W., 1989. Schiitzung von Parametern der Leistung holl~indischer Warmblutpferde in Spring- und Dressur-Wettbewerben. Livest. Prod. Sci. 21:333-345 (auf englisch). Ziel der Arbeit ist die Sch~itzung verschiedener genetischer Parameter in der holl~indischen Warmblut-Reitpferde°Population.Die berticksichtigten Merkmale sind die Leistungen in Springund Dressur-Wettbewerben. Folgende Parameter wurden geschiitzt: Heritabilit~it for die Spring- und Dressurleistung und genetische Korrelationen zwischen diesen, phiinotypische und genetische Korrelationen zwischen den Leistungen bei verschiedenem Alter. Ftir die Parameter-Schiitzung wurde die Methode "Restricted Maximum Likelihood (REML)" angewandt.
345 Es standen Daten von 6899 Pferden mit Springleistungen und 10 408 Pferden mit Dressurergebnissen zur Verfiigung, die von 205 bzw. 237 Hengsten abstammen. Das Alter der Nachkommen variierte zwischen 4 und 8 Jahren. Als Leistungsergebnis wurde eine kumulativ abgeleitete Note verwendet, die das Niveau der Leistung im Wettbewerb widerspiegelt. Eine Quadratwurzel-Transformation wurde verwendet, um eine Normalverteilung der Daten zu erreichen. Fiir die Schiitzung der phiinotypischen und genetischen Parameter wurden die Daten in zwei Datens~itze entsprechend des Alters der V~iter aufgeteilt. Ftir die Beziehung zwischen Leistungen bei 4, 5 und 6 Jahren wurden die Leistungen der Nachkommen friiherer Jahre mit den vorliegenden Daten verkniipft. Die unverzerrtesten Heritabilit~itsschiitzwerteflit die Spring- und Dressurleistung ergaben sich ftir die jiingsten Nachkommen der jiingsten Hengste mit 0,20 bzw. 0,10. Die genetische Korrelation zwischen der Spring- und Dressurleistung lag im Bereich v o n - 0,27 und + 0,10. Die ph~notypische Korrelation zwischen diesen Merkmalen schwankte zwischen + 0,15 und + 0,26. Phiinotypische und genetische Korrelationen zwischen den Leistungen bei 4, 5 und 6 Jahren lagen im Mittel bei 0,95 bwz. 0,75. Die zuletzt aufgefiihrten Ergebnisse haben bedeutende Auswirkungen auf die Zuchtwertschiitzung in der Population.