REGIONAL HETEROGENEITY OF VARIANCES AND ITS EFFECT ON CANADIAN HOLSTEIN SIRE EVALUATION

Abstract

Heritabilities were estimated within each of four artificial insemination stud regions in Ontario and Quebec, based on a total of 880 329 records for milk and fat yield and 441 230 records for final class score. The pseudo-expectation (PE) and tilde-hat (TH) approaches were used. Estimates ranged from 0.32 to 0.46 for milk yield, from 0.31 to 0.53 for fat yield, and from 0.17 to 0.47 for final class score. Differences between methods (PE and TH) were as large as differences between regions for all traits. TH regional estimates were less variable than PE estimates. Changes in sire rankings were trivial when using regional estimates versus using a common variance ratio in each region during evaluation. Rank correlations were 0.98 and higher between evaluations based on heterogeneous regional variances and evaluations based on homogeneous variances. For all three traits, variance ratios that have previously been assumed for national evaluation were more appropriate than estimates obtained in this study. Use of the previously assumed values resulted in a greater consistency, measured by the correlation, between the regional proofs of a sire and between a sire's proof and pedigree index. Key words: Pseudo-expectation, tilde-hat, heterogeneous variances, Holstein, sire evaluation