Comparative efficiency of different multiple linear regression prediction equations of first lactation 305-day milk yield for sire evaluation in Murrah buffaloes

Abstract

In this study, first lactation 39059 weekly test-day milk yield records of 961 Murrah buffaloes were used to predict first lactation 305-day milk yield (FL305DMY) by stepwise backward regression method. The best single, two, three and four test day combinations were selected for prediction of FL305DMY based on adjusted R2 and RMSE values. The sires were evaluated for 305-day actual and predicted first lactation milk yield based on derived multiple regression equations using four methods viz. least squares (LSQ), simple regressed least squares (SRLS), best linear unbiased prediction sire model (BLUP-SM) and best linear unbiased prediction animal model (BLUP-AM) methods. The effectiveness of different sire evaluation methods were judged by error variance, coefficient of determination, coefficient of variation and spearman’s rank correlation. The accuracy of prediction of FL305DMY from weekly test day milk yields were observed to be best for TD-7 (48th day) and TD-22 (153rd day) combination with BLUP-AM as the most efficient method for sire evaluation. It was concluded that the FL305DMY can be predicted as early as 153rd day of lactation and further can be used for early genetic evaluation of Murrah sires.