Prediction of genetic parameters related to milk yields of holstein friesian population reared in the Ceylanpınar Agricultural State Farm by bayesian and REML methods

In the present study, 4807 lactation milk yield records obtained from Holstein Friesian population reared in the Ceylanpinar Agricultural State Farm between 1990 and 1997 were used for the estimation of variance components, heritability and breeding values of animals by Bayesian and Restricted Maximum Likelihood (REML) methods. REML estimates were obtained using both the animal (with DFREML Ver. 3.0 P) and the sire models (with SAS Institute 1987, Release 6.03). In the analyses, year-season was assumed to be a fixed effect and calving age and lactation number were covariates. REML estimates of heritability for lactation milk yield using the sire model and using the animal model were 0.28 and 0.43 respectively. Bayesian estimates were obtained using Gibbs sampling and heritability was calculated as 0.27. Analyses showed that year-season, lactation number and age at calving (linear) significantly affected 305-day milk yield (p<0.01, p<0.05 and p<0.01 respectively). When sires were ranked according to estimated breeding values, REML and Bayesian estimates using the sire model were almost identical (Kendall's tau correlation coefficient was 0.997). However, REML estimates using the animal model by DFREML were different (Kendall's tau correlation coefficients were between 0.744 and 0.745) from the other methods. In addition, the rank correlation between estimated breeding values and mean milk yields of cows was quite low (Kendall's tau correlation coefficient was 0.377).Key Words: REML, Bayesian, Holstein Friesian, Genetic Parameter, Milk Yield