Estimação de Componentes de (co)variância para Características de Crescimento em Bovinos da Raça Nelore Criados na Região do Trópico Úmido do Brasil por Meio de Regressão Aleatória.

Abstract

This study aimed to estimate (co)variance components and genetic parameters for growth traits using random regression models using the condition of homogeneity and heterogeneity of residual variance. Information of animals from birth to adulthood (born between 1993 and 2010) of Nelore cattle were used. The data were provided by the National Association of Breeders and Researchers (ANCP), from animals raised in the Humid Tropics of Brazil, belonging to the states of Mato Grosso, Pará, Rondônia and Tocantins. We used 23,278 records of weighings. The model as fixed effect of contemporary group, and as covariates age at calving (linear and quadratic effect) and the date the animal at weighing (linear and quadratic effect), and as random genetic additive direct and maternal effects and permanent environmental animal and maternal. The residue was modeled using seven classes of variances, grouped as follows: 1-56, 57-147, 148-218, 219-259, 260-301, 302-372 and 373-550 days of age, respectively. Considered orthogonal Legendre polynomial, cubic order on age weighing model, to model the mean curve of the population. Sixteen models of orthogonal Legendre polynomials ranging from first to sixth order were used to describe the additive genetic direct and maternal, permanent environmental and animal nursery. The different models were compared by the Akaike information criteria (AIC) and Schwarz Bayesian (BIC). Heritability estimates for direct genetic effect is presented increasing from 120 to 450 days of age and were reduced to 550 days of age, observing values of 0.23, 0.38, 0.52 and 0.43, respectively, and the estimates of maternal heritability were low for all ages analyzed, ranging from 0.01 to 0.08. The genetic correlations were moderate to high magnitude, and remaining even with the moderate increase of the distance between the ages. The use of models that considered the heterogeneity of residual variance were more precise and accurate in the estimation and prediction of the (co) variance.