ASSESSING THE EFFICIENCY OF MULTIPLICATIVE MIXED-MODEL EQUATIONS TO ACCOUNT FOR HETEROGENEOUS VARIANCE ACROSS HERDS IN CARCASS SCAN TRAITS FROM BEEF-CATTLE

Data (n = 2,658) from live animal ultrasonic measures from 17 Angus he rds were used to evaluate a multiplicative mixed model that incorporat es scaling factors to correct for across-herd heterogeneity of varianc e. Traits included were ribeye muscle area (EMA), surface fat at the P 8 site (P8), surface fat between the 12th and 13th ribs (RIB12), and w eight at scanning (WEIGHT). Cattle ranged in age from 501 to 698 d and represented 291 contemporary groups. Data were initially analyzed usi ng single-trait, animal model, Method R procedures to estimate varianc e components and heritabilities (h(2)). These estimates were incorpora ted into a multiplicative mixed model that simultaneously estimates br eeding values (EBV) and heterogeneity factors. Reestimation of h(2) af ter scaling the data with the some traits in the smaller herds. correc tion factors was explored to obtain a measure of the improvement in th e genetic evaluation and to detect changes in ranking of individuals a nd herds. Initial h(2) estimates for EMA, P8, RIB12, and WEIGHT were . 36, .39, .29, and .48, respectively. Scaling factors ranged from .25 f or P8 in a herd with eight records to 1.96 for RIB12 in a herd with 86 individuals. Re-estimates of h(2) increased by an average of 4.2% for all the traits as a result of correcting for heterogeneity. Deviation s of new scaling factors were within expectations. Correlations betwee n EBV with and without heterogeneity correction were greater than .97 for all the traits. However, some substantial re-rankings of herds wer e observed for some traits in the smaller herds.