The comparison of alternative models for genetic evaluation of growth traits in Lori-Bakhtiari sheep: Implications on predictive ability and ranking of animals

Abstract

The aim of present study was to investigate the performance of standard multivariate model (SMM), fully recursive models (FRM) and temporal recursive model (TRM) for genetic evaluation of growth traits in Lori-Bakhtiari sheep. The pedigree and phenotypic data of growth traits including birth weight (BW), average daily gain from birth to weaning (ADG1), weaning weight (WW), average daily gain from weaning to six-month weight (ADG2) and six-month weight (6 MW) collected from 1995 to 2012 were used. Firstly, three models (SMM, FRM and TRM) were fitted via Bayesian approach with 200,000 Gibbs samples and the first 50,000 samples were considered as burn-in period with thinning intervals of 10 samples. Contrary to FRM and TRM, in SMM causal relationships between the studied traits were ignored. Deviance Information Criterion (DIC) values obtained under three considered models indicated superiority of models with causal relationships on SMM. Also, based on DIC, within models containing causal relationships, TRM performed better than FRM for genetic evaluation of the studied growth traits. The posterior means for structural coefficients between BW-ADG1, ADG1-WW, WW-ADG2 and ADG2-6 MW were 9.343, 0.03, 10.632 and 0.14, respectively and were statistically significant (P < 0.05). Results of comparisons of rank correlations between posterior means of direct genetic effects for the studied growth traits under SMM and TRM revealed that taking the causal relationships among the studied growth traits into account may cause considerable re-ranking for the animals in terms of the estimated breeding values, especially for the top-ranked animals. Model comparisons based on goodness of fit and predictive ability applying mean square error (MSE) and Pearson’s correlation coefficients between the observed and predicted records (ry,yˆ(revealed that in general TRM performed better than SMM in terms of goodness of fit and predictive ability. In general, the results indicated that considering causal relationships between growth traits is very important and including these relationships in models for genetic evaluation is beneficial in terms of predictive ability.