Prediction of first lactation milk yield using multiple regression analysis and principal components analysis in Ethiopian Boran and their crosses with Holstein Friesian in Central Ethiopia

Abstract

The data on 60 Ethiopian Boran and 428 Boran‐Friesian crosses spread over 15 years (1990‐ 2004) were used to predict first lactation milk yield (FLYD) based on earlier expressed traits using multiple regression analysis and principal components analysis. In the prediction of FLYD in Ethiopian Boran cattle, the step‐wise regression found only AFCONC to be significantly associated with FLYD (R2=13.6 percent) in adjusted data. In the Boran‐Friesian crosses, Gain 2, BWT, AFS and YWT were significantly associated with FLYD (R2= 15.21 percent). The first principal component (PC1) out of 4 fitted composite variables was significantly associated with FLYD in Ethiopian Boran breed (R2= 13.29 percent). The third principal component (PC3) was more related to FLYD in Boran‐Friesian crosses (R2= 11.83 percent) and inclusion of next important principal component PC1 improved accuracy of prediction to 13.76 percent. The rest of the components were significant and dropped. The results indicated that first lactation milk yield could be predicted directly from the early expressed growth and reproductive performance traits, though with lower accuracy, and transformation of original variables into principal components offers no additional advantage in terms of accuracy of prediction.