Variable selection is one of the most fundamental problems in regression analysis. By sampling from the posterior distributions of candidate models, Bayesian variable selection via MCMC （Markov chain Monte-Carlo） is...Variable selection is one of the most fundamental problems in regression analysis. By sampling from the posterior distributions of candidate models, Bayesian variable selection via MCMC （Markov chain Monte-Carlo） is effective to overcome the computational burden of all-subset variable selection approaches. However, the convergence of the MCMC is often hard to determine and one is often not sure about if obtained samples are unbiased. This complication has limited the application of Bayesian variable selection in practice. Based on the idea of CFTP （coupling from the past）, perfect sampling schemes have been developed to obtain independent samples from the posterior distribution for a variety of problems. Here the authors propose an efficient and effective perfect sampling algorithm for Bayesian variable selection of linear regression models, which independently and identically sample from the posterior distribution of the model space and can efficiently handle thousands of variables. The effectiveness of the authors＇ algorithm is illustrated by three simulation studies, which have up to thousands of variables, the authors＇ method is further illustrated in SNPs （single nucleotide polymorphisms） association study among RA （rheumatoid arthritis） patients.展开更多
During the past decade,shrinkage priors have received much attention in Bayesian analysis of high-dimensional data.This paper establishes the posterior consistency for high-dimensional linear regression with a class o...During the past decade,shrinkage priors have received much attention in Bayesian analysis of high-dimensional data.This paper establishes the posterior consistency for high-dimensional linear regression with a class of shrinkage priors,which has a heavy and flat tail and allocates a sufficiently large probability mass in a very small neighborhood of zero.While enjoying its efficiency in posterior simulations,the shrinkage prior can lead to a nearly optimal posterior contraction rate and the variable selection consistency as the spike-and-slab prior.Our numerical results show that under the posterior consistency,Bayesian methods can yield much better results in variable selection than the regularization methods such as LASSO and SCAD.This paper also establishes a BvM-type result,which leads to a convenient way of uncertainty quantification for regression coefficient estimates.展开更多
文摘Variable selection is one of the most fundamental problems in regression analysis. By sampling from the posterior distributions of candidate models, Bayesian variable selection via MCMC （Markov chain Monte-Carlo） is effective to overcome the computational burden of all-subset variable selection approaches. However, the convergence of the MCMC is often hard to determine and one is often not sure about if obtained samples are unbiased. This complication has limited the application of Bayesian variable selection in practice. Based on the idea of CFTP （coupling from the past）, perfect sampling schemes have been developed to obtain independent samples from the posterior distribution for a variety of problems. Here the authors propose an efficient and effective perfect sampling algorithm for Bayesian variable selection of linear regression models, which independently and identically sample from the posterior distribution of the model space and can efficiently handle thousands of variables. The effectiveness of the authors＇ algorithm is illustrated by three simulation studies, which have up to thousands of variables, the authors＇ method is further illustrated in SNPs （single nucleotide polymorphisms） association study among RA （rheumatoid arthritis） patients.
基金supported by National Science Foundation of USA(Grant No.DMS1811812)supported by National Science Foundation of USA(Grant No.DMS-2015498)National Institutes of Health of USA(Grant Nos.R01GM117597 and R01GM126089)。
文摘During the past decade,shrinkage priors have received much attention in Bayesian analysis of high-dimensional data.This paper establishes the posterior consistency for high-dimensional linear regression with a class of shrinkage priors,which has a heavy and flat tail and allocates a sufficiently large probability mass in a very small neighborhood of zero.While enjoying its efficiency in posterior simulations,the shrinkage prior can lead to a nearly optimal posterior contraction rate and the variable selection consistency as the spike-and-slab prior.Our numerical results show that under the posterior consistency,Bayesian methods can yield much better results in variable selection than the regularization methods such as LASSO and SCAD.This paper also establishes a BvM-type result,which leads to a convenient way of uncertainty quantification for regression coefficient estimates.
