Prediction error variances for interbreed genetic evaluations

Abstract

A table for adjusting expected progeny differences (EPD) to a base year and breed basis depends on analyses of records of progeny of bulls of different breeds in a common environment and requires that those reference bulls also have other progeny to provide within-breed EPD. Currently, the germ plasm evaluation project at the Meat Animal Research Center (MARC) provides such a common environment for reference bulls of several breeds for estimation of breed differences for the reference sires. Reference sire estimates of breed differences are adjusted by the difference between average EPD of reference bulls and average EPD for the base year for that breed. Two related questions are as follows: 1) What are confidence ranges for the adjustments and 2) What are accuracies of interbreed EPD? Application of statistical principles and algebra shows that 1) apparent confidence ranges for breed adjustments are small, 2) apparent confidence ranges are substantially underestimated when random sire effects within breed are ignored, 3) correct confidence ranges also are small, 4) usual measures of accuracy cannot be applied to interbreed comparisons, and 5) standard errors of prediction used in calculating confidence ranges for interbreed comparisons are much less affected by variance of the adjustment factors than by within-breed accuracies for two bulls being compared except for bulls with accuracies of near unity. Alternatives of predicting differences between bulls of the same or different breeds or between a bull of any breed and an average bull of a base breed are discussed in terms of confidence ranges.(ABSTRACT TRUNCATED AT 250 WORDS)