B1829
Title: Strategies for high-dimensional empirical Bayes problems
Authors: Sihai Dave Zhao - University of Illinois Urbana-Champaign (United States) [presenting]
Abstract: Many modern data analysis problems require simultaneous estimation and/or inference for a large number of features. These problems are amenable to empirical Bayes approaches, which share information across the features. Nonparametric empirical Bayes methods are especially useful because they can automatically identify the optimal way to share that information in a given dataset. However, when the parameters of interest are of moderate or high dimension, nonparametric methods suffer from the curse of dimensionality and become extremely inaccurate. There are few practical solutions. Some motivating problems are introduced, from the fields of metabolomics and spatial transcriptomics, and new strategies are then explored for overcoming dimensionality when using nonparametric empirical Bayes methods.
