Bayesian segmented regression model for adaptability and stability evaluation of cotton genotypes

Abstract

The genotype by environment interaction is essential in any plant breeding program. Methodologies allowing the evaluation of nonlinear genotype responses to environmental variation allied to prior beliefs on unknown parameters bring new insights for breeders. In this context, we aimed to propose a Bayesian segmented regression model to infer on phenotypic adaptability and stability of cotton (Gossypium L.) cultivars. The efficiency of using informative and minimally informative prior distributions in the selection of cultivars was also investigated. Randomized complete-block design experiments to study fiber yield (kg/ha) of 16 cotton genotypes were carried out in eight different environments in the State of Mato Grosso, Brazil. The proposed methodology was implemented using the free software R through the rbugs package. Bayesian segmented regression model was able to recommend cotton genotypes for cultivation, allowing to exploit the nonlinear pattern of genotype responses to environmental variation to find out the “ideal” genotype. The use of suitable prior information reduces the ranges of the credibility intervals, implying in higher precision of parameter estimates and, consequently, in reliable genotypes selection.