Multivariate, Heteroscedastic Empirical Bayes via Nonparametric Maximum Likelihood

TL;DR

This paper develops a nonparametric empirical Bayes method for multivariate, heteroscedastic denoising problems, providing theoretical guarantees and demonstrating practical effectiveness in astronomy, education, and genomics.

Contribution

It extends the NPMLE framework to handle multivariate, heteroscedastic errors and proves its near-optimality and adaptability in various applications.

Findings

Low regret of empirical Bayes posterior means

Nearly optimal denoising performance with finite-sample bounds

Successful application to astronomy, education, and genomics datasets

Abstract

Multivariate, heteroscedastic errors complicate statistical inference in many large-scale denoising problems. Empirical Bayes is attractive in such settings, but standard parametric approaches rest on assumptions about the form of the prior distribution which can be hard to justify and which introduce unnecessary tuning parameters. We extend the nonparametric maximum likelihood estimator (NPMLE) for Gaussian location mixture densities to allow for multivariate, heteroscedastic errors. NPMLEs estimate an arbitrary prior by solving an infinite-dimensional, convex optimization problem; we show that this convex optimization problem can be tractably approximated by a finite-dimensional version. The empirical Bayes posterior means based on an NPMLE have low regret, meaning they closely target the oracle posterior means one would compute with the true prior in hand. We prove an oracle inequality implying that the empirical Bayes estimator performs at nearly the optimal level (up to logarithmic factors) for denoising without prior knowledge. We provide finite-sample bounds on the average Hellinger accuracy of an NPMLE for estimating the marginal densities of the observations. We also demonstrate the adaptive and nearly-optimal properties of NPMLEs for deconvolution. We apply our method to two denoising problems in astronomy, constructing a fully data-driven color-magnitude diagram of 1.4 million stars in the Milky Way and investigating the distribution of 19 chemical abundance ratios for 27 thousand stars in the red clump. We also apply our method to hierarchical linear models, illustrating the advantages of nonparametric shrinkage of regression coefficients on an education data set and on a microarray data set.