Bayesian inference describing statistical models based on the average daily gain of Nile Tilapia (<em>Oreochromis niloticus</em> “GIFT”)

Abstract

The aim of the current study was to analyze four statistical models using Bayesian inference to describe the average daily gain (g) (ADG) during the genetic selection of the Nile Tilapia (Oreochromis niloticus “GIFT”). The data set had records from 2,615 fish collected from the fourth generation (G4) of the breeding programme in the Floriano Breeding Station of the Universidade Estadual de Maringá, Maringa County, Paraná State, Brazil. In these analyses, we considered an animal model where sex was the fixed effect, linear and quadratic effects of the fish age was covariates in days in conjunction with the additive genetic effects. The models were modified based on the available information from additive genetic effects and common environments of hatchery (c), nursery (w), or none of them. The heritability results were estimated for the models, M2=0.24 and M4=0.23 and high for M1=0.83 and M3=0.79. The criterion of selection in the DIC model was lower for the M1 with -282.59 and higher for M4 with 1754.57. Another criterion of selection was the marginal log density of the Bayer factor which corroborates with the DIC only for the lower value in which M1=585.29. Less computer efforts to achieve convergence was found using the M1 model with 15,000 chains, which was the best model to explain and predict the phenomenon.