Abordagem bayesiana e freqüentista em análise genética tricaráter para crescimento e reprodução de bovinos nelore

Abstract

The Bayesian inference has been proposed as an alternative for estimating variance components instead of the frequentist approach. The objective of this study was to estimate the (co)variance components and genetic parameters of growth and reproduction traits in Nelore cattle using three-trait animal model by Restricted Maximum Likelihood and Gibbs Sampling. The data set had 15,173 and 6,911 records of weight (W365) and scrotal circumference (SC365) at 365 days of age, respectively, and 10, 388 records of age at first calving (AFC), from the Nelore Breed Genetic Improvement Program farms. The linear model included the contemporary groups and age-of-dam (with exception for AFC) as fixed effects, and residual and additive direct genetic effects as random. Estimates obtained by Restricted Maximum Likelihood were different from those found by Gibbs Sampling. There were no differences in genetic parameter estimates using the three levels of prior information by Gibbs Sampling. The Bayesian inference has advantages in relation to Frequentist approach due to marginal distributions which offer more information about the parameters.