Title: Nonparametric empirical Bayes methods for sparse, noisy signals
Authors: Junhui Jeffrey Cai - University of Pennsylvania (United States) [presenting]
Linda Zhao - University of Pennsylvania (United States)
Abstract: High dimensional signal recovering problems will be considered. The goal is to identify the true signals from the noise controlling false discovery rate and then to make inference for the unknown signals. We propose a nonparametric empirical Bayesian scheme to tackle the problem. The method adapts well to varying degrees of sparsity. It not only performs well to recover the signals, but also provides credible intervals. The method is built upon noisy data with exponential family distribution. It covers large range of data structure such as normal means with heteroskedastic variance, Poisson data with varying degrees of frequency, and Binomial counts. Simulations show that our method outperforms existing ones. Applications in microarray data as well as sport data such as predicting batting averages will be discussed.