Consistency of modified versions of Bayesian Information Criterion in sparse linear regression with subgaussian errors

Abstract

We consider a sparse linear regression model, when the number of available predictors, $p$, is much larger than the sample size, $n$, and the number of non-zero coefficients, $p_0$, is small. To choose the regression model in this situation, we cannot use classical model selection criteria. In recent years, special methods have been proposed to deal with this type of problem, for example modified versions of Bayesian Information Criterion, like mBIC or mBIC2. It was shown that these criteria are consistent under the assumption that both $n$ and $p$ as well as $p_0$ tend to infinity and the error term is normally distributed. In this article we prove the consistency of mBIC and mBIC2 under the assumption that the error term is a subgaussian random variable.