Comparison of two different statistical methods in estimation of (co)variance components of milk production traits of Iranian Holstein cows

The aim of this study was comparison of restricted maximum likelihood and Bayesian statistical methods to estimate of (co)variance components of production traits of Iranian Holstein cows by multi-trait random regression animal model. Data included 181,226 test-day records collected on 21,469 fist lactation cows. The model included herd-year-season of calving, year-month of test day, somatic cell score, age of calving and milking time as fixed effects, and random regression coefficients for additive genetic, and permanent environmental effects. For restricted maximum likelihood method, heritability's of test-day milk yield were higher in the middle (0.30) than at the beginning (0.08) or the end (0.23) of lactation, but for Bayesian method the highest value was observed for end of lactation (0.31). Genetic correlations among 305-d yield traits ranged from -0.502 (milk yield-protein percentage) to 0.712 (fat-protein percentage) for restricted maximum likelihood method and from -0.525 (milk yield-protein percentage) to 0.703 (fat-protein percentage) for Bayesian method. Residual variances of production traits for first method were 0.005 (%2) to 0.64 (kg2) higher than these values for second method.