Using the Bayesian Model Averaging Approach for Genomic Selection by Considering Skewed Error Distributions

Background and Purpose: Genomic selection is used to select candidates for breeding programs for organisms. In this study, we use the Bayesian model averaging (BMA) method for genomic selection by considering the skewed error distributions. Materials and Methods: In this study, we apply the BMA method to linear regression models with skew-normal and skew-t distributions to determine the best subset of predictors. Occam’s window and Markov-Chain Monte Carlo model composition (MC3) were used to determine the best model and its uncertainty. The Rice SNP-seek database was used to obtain real data, which included 152 single nucleotide polymorphisms (SNPs) with 6 phenotypes. Results: Numerical studies on simulated and real data showed that, although Occam’s window ran faster than the MC3 method, the latter method suggested better linear models for the data with both skew-normal and skew-t error distributions. Conclusion: The MC3 method performs better than Occam’s window in identifying the linear models with greater accuracy when dealing with skewed error distributions.