Inferência sobre gráficos de radar: um exemplo no melhoramento vegetal

Abstract

This study aimed to develop a tool for multivariate selection based on Bayesian inference from each trait to identify superior genotypes in plant breeding programs, using the radar chart to assess the regularity and balance of genotypes, highlighting both performance consistency and advantageous specializations. Initially, a simulated experiment evaluated the methodology’s potential, which was later applied to a real study involving ten agronomic and technological traits in 45 sweet sorghum genotypes. The proposed technique can be adjusted according to breeders’ interests, making it applicable to other crops and genotype-environment interaction studies. Although the Bayesian approach is flexible and provides complete distributions for the parameters, similar results could be obtained using REML estimates, or similar techniques from mixed models. In addition to facilitating selection based on multiple traits, the proposed method is easy to apply and shows great potential in plant breeding.