Accuracy of different methods of sire evaluation for production efficiency trait in Sahiwal cattle

Abstract

Thc data were collected from 380 Sahiwal cows, progeny of 26 sires maintained at the Government Livestock Farm, during 20 years (1974-93) were used to estimate first lactation milk yield/day of age at second calving and to compare the sire's breeding values estimated by ordinary least-squares (OLS), regressed least-squares (RLS) and best linear unbiased prediction (BLUP). The same model was used for all 3 proccdures, which included period and, season of calving as fixed effects, age at first calving as covariable and sire as random effect. Rank and product moment correlations among estimated sire merit obtained from OLS and RLS methods were high and significant. OLS and RLS estimates of breeding value had very low and nonsignificant association with estimates of breeding values obtained from BLUP. RLS cannot be recommended because of computational difficulty though coefficient of skewness and kurtosis obtained from the estimates of breeding value were close to normal distribution. It is recommended to use BLUP procedure in a situation where correct ratio of residual to sire variances in known. Use of OLS is recommended in a situation where correct ratio of residual to sire variances is unknown.