Impact of Bayesian Inference on the Selection of Psidium guajava

Abstract

Perennial breeding species demand substantial investment in various resources, mainly the required time to obtain adult and productive plants. Estimating several genetic parameters in these species, in a more confidence way, means saving resources when selecting a new genotype. A model using the Bayesian approach was compared with the frequentist methodology for selecting superior genotypes. A population of 17 families of full-siblings of guava tree was evaluated, and the yield, fruit mass, and pulp mass were measured. The Bayesian methodology suggest more accurate estimates of variance components, as well as better results to fit of model in a cross-validation. Proper priori for Bayesian model is very important to convergency of chains, mainly for small datasets. Even with poor priori, Bayesian was better than frequentist approach.