Comparison of alternative models for genetic evaluation of growth traits of registered Brahman cattle in Mexico

Abstract

Five genetic evaluation models of registered Brahman cattle in Mexico were compared for birth weight (PN), weaning (PD), yearling (PA) and final weight at 550 days (PF), and their effect on genetic parameter estimation and EPD ordering were also assessed. Models differed on number of random factors; D, model with direct genetic effect (σ2d), DM, model like D plus maternal genetic effect (σ2m) and genetic covariance between genetic effects equal to zero (σdm = 0), DP, D like model plus maternal permanent effect (σ2p), DMC, DM like model plus covariance between genetic effects (σdm ≠ 0), and DMCP as the complete model, equal to DMC plus (σ2p). Models included the fixed effect of contemporary grouping (herd, sex, year and season of birth), and age of dam as linear and quadratic covariable. A likelihood ratio test (PRV) was performed for model comparison, and EPD reordering was quantified by Spearman rank correlation analysis. The PRV showed DMC as the best model for all variables, except in PD. For PD the best model was DP. A negative genetic correlation rdm (rdm = σdm2) was observed for PN, PA and PF (-0,86, -0,84 and -0,52, respectively) reducing the magnitude of total heritability estimates. For PD, DMC and DMCP did not converge. Correlation between selected models and the other compared models, suggested that there might be important and substantial changes on 10% of superior animals. Structure of data may modify the genetic parameter estimation if adjusting complex models; hence the model selection could be crucial before performing any genetic evaluation. Considerations were made concerning selective recording.