Comparison of methods for estimating components of variance and related parameters of multiple traits in cattle

Abstract

The objective of this study vas to compare Maximum Likelihood (ML), Restricted Maximum Likelihood (REML) and the Iterative Henderson Simple Method (IHSM), regarding computational aspects, genetic (g_ii), residual (r_ii) variance and heritability (h²) estimates from a multitrait model of body weight at weaning and at 12 months of age in cattle. IHSM, ML and REML, converged with 13, 40 and 40 iterations, respectivelly. Compared to ML and REML, IHSM overestimated g_ii and h² values. Although iterative, the IHSM method required less calculations and appeared attractive in estimating (co) variances from multiple-trait models in large populations including relationship among animals. However, additional investigations are needed, before its recommendation for practical use. Compared to REML, ML underestimated values of r_ii and overestimated the h² values in 5%. Due to computational task for using REML, ML can be useful in estimating (co) variances from multiple-trait model, including relationship among sires.