Comparison of Henderson's Method I and restricted maximum likelihood estimation of genetic parameters of reproductive traits.

Genetic parameters of simulated reproductive traits were estimated with Henderson's Method I or restricted maximum likelihood (REML). Each of the 25 samples consisted of 2,667 birds from 50 sires. When reproductive traits were measured on all birds, the means of both sets of estimates were equal to the true value; however, precision was higher with REML. When they were only measured on the 60 (20%) heaviest birds, Henderson's Method I was severely biased by selection (up to 40% for estimates of heritability of reproductive traits and up to 45% for estimates of genetic correlation between reproductive traits and body weight). This bias, as well as the variance of estimates, increased with selection rate and absolute value of genetic correlation between body weight and reproductive traits (or another criterion of selection). Conversely, REML was not biased by selection. It should therefore be preferred when sequential selection occurs.