Mixed models are suited to describe the parameterization needed to estimate variance components due to genotypes, the environment and genotype × environment interaction over several locations and years. In Bayesian approach, incorporating the prior information of variance component from multi environment trials on the genotypic parameters available from previous similar trials has potential for adding value to the crop breeding program and genetic variability. The objective of this study was to obtain Bayesian estimates of variance components, heritability in broad-sense and genetic advance due to selection for seed yield of chickpea. Chickpea yield (kg/ha) on twelve genotypes data were collected from a series of multi-year multi-location trials conducted in randomized complete block designs in Indian environments. An MCMC estimator is implemented in the WinBUGS and R software for Bayesian posterior. The differences in variance component estimates obtained by two approaches, the classical approach using restricted maximum likelihood method and the Bayesian approach, were investigated. Bayesian estimate of heritability for seed yield on the plot-basis was different from that on the mean-basis, as may be expected. For seed yield, the Bayesian estimates of heritability were 9% on plot basis and 52% on mean basis, and the genetic advance due to selection was 7% using half-t prior. and were 13% on plot-basis and 58% on mean-basis, and the genetic advance due to selection was 8% using half-normal prior, which is higher in comparison to the frequentist approach.
展开▼