Monte Carlo Comparison of Sire Evaluation Models in Populations Subject to Selection and Nonrandom Mating

Abstract

Dairy populations of constant size (2000 cows) were characterized by culling of females for voluntary and involuntary reasons, progeny testing of males, and selection on predicted transmitting abilities. Three simulated mating schemes were 1) random culling and mating (control); 2) active selection with random mating; 3) active selection with mating of likes. Predicted transmitting abilities of males were calculated from a) individual animal model, b) sire model, c) sire and maternal grandsire model, and d) sire model with dam predicted transmitting ability as covariate. Accuracy was square root of mean squared differences between true and predicted transmitting abilities. In control replicates, accuracy of predicted transmitting abilities were similar among models. In actively selected, random mating replicates models b, c, and d were 43 to 47% less accurate than model a. In actively selected, like mating populations model b was 65 to 67% less accurate than model a, and models c and d were 56 to 60% less accurate than model a. Sires related increased the accuracy of models b, c, and d. Including maternal grandsire or dam predicted transmitting ability in the sire model partially accounted for bias in sire predicted transmitting abilities from mating of likes.