مقايسه كارآيي دو مدل ساختار واريانس باقي مانده صفت توليد شير بر اساس تجزيه ركوردهاي روزآزمون درمدل هاي رگرسيون تصادفي

Comparison Structure of residual variance of milk yield trait based on the analysis of test day records in random regression models

With analyzing test day milk records of first lactation Holsteins using random regression models, the effect of assuming a constant residual variance during lactation on the other (co)variance components and heritability of milk yield was studied. Additive genetic and permanent environmental covariance functions were estimated using 3rd order orthogonal Legendre polynomial and different assumptions of residual variance in the analysis including the assumption of constant (homogenous) and variable (heterogeneous) residual variance during lactation were investigated. According to the results, the assumption of variable residual variance during lactation resulted in significant increase (P<0.05) in the log-likelihood and improved the fitness of derived covariance functions. Maximum residual variance of milk yield was in early lactation, then declined to mid lactation rapidly and finally increased a little in late lactation. Assuming a constant residual variance during lactation, residual variance was underestimated and heritability was overestimated nearly in mid lactation. Because the number of estimated parameters increases with the assumption of variable residual variance, using of a variance function for different estimates of mentioned component for each day or stage of lactation was recognized as suitable.