- Comparison of breeding values for body weight in Merino sheep estimated with different statistical procedures
Comparison of breeding values for body weight in Merino sheep estimated with different statistical procedures
K.R. Nemutandani1#, M.A. Snyman1, W.J. Olivier1 & C. Visser2
1Grootfontein Agricultural Development Institute, Private Bag X529, Middelburg (EC), 5900, South Africa
2Department of Animal & Wildlife Sciences, Faculty of Natural & Agricultural Sciences, University of Pretoria, 0002, South Africa
# Corresponding author: Khetho Nemutandani
Background: Genomic selection has revolutionised animal breeding. However, before genomic selection can be implemented, it is imperative that accurate and reliable conventional estimated breeding values (EBV) for performance traits are available. It is therefore important that EBVs for performance traits should be estimated as accurately as possible. This could be achieved by fitting the most appropriate model, which accounts for all known non-genetic effects, as well as correctly partitions the genetic variance into its various sources.
Aim: The aim of this study was to determine the most suitable statistical procedures for estimation of breeding values for body weight in Merino sheep for inclusion in a genomic selection program.
Methodologies: Data for body weight recorded at different ages in the Grootfontein Merino stud from 1968 to 2012 were used for this study. Various statistical procedures, including univariate and bivariate linear models employing restricted maximum likelihood methods, repeatability models and random regression procedures were used to estimate (co)variance components, genetic parameters and breeding values for the different body weights. Estimated breeding values for weaning weight (WW), weight at 15 months (W15) and 3-year old body weight (BW3) obtained with the most suitable univariate and bivariate models, as well as the most suitable repeatability model, were compared. The Spearman ranking correlation option under the PROC CORR-procedure of SAS was used to estimate the correlations between the ranks of these EBVs. Furthermore, lists of the top 10% of all animals, animals with records and sires were compiled for EBVs estimated with the different procedures. The number and percentages of animals common to the different lists were compared.
Results: Significant Spearman rank correlations of 0.87, 0.97 and 0.55 were estimated between EBVs obtained with univariate and bivariate animal models for WW, W15 and BW3 respectively. Comparing EBVs estimated with a repeatability model with uni- and bivariate model EBVs for WW, yielded Spearman correlations of 0.24 and 0.25. Corresponding Spearman rank correlations of 0.22 and 0.17 were obtained for W15 and 0.10 and 0.18 for BW3. The number of common animals in the top 100 lists for uni- and bivariate EBVs was 38 for WW, 83 for W15 and 48 for BW3. When comparing the top 100 list for repeatability model EBVs with those of uni- and bivariate EBVs, the number of common animals ranged from 2 (W15) to 17 (BW3).
Discussion: Ranking of animals on EBVs differed considerably when comparing repeatability model EBVs with those of uni- and bivariate EBVs. The highest correlations were obtained for WW and W15. The most complete data sets were available for these two weights, which could have contributed to more accurate EBVs being estimated for these weights.
Conclusions: Estimation of EBVs with random regression and multivariate models are underway and these will also be included in the comparison, in order to determine the most accurate procedures for estimation of breeding values for Merino sheep for ultimate inclusion in genomic selection.
Proceedings 49th SASAS congress, Stellenbosch