Bayesian Tapering Test for Comparing Two Estimated Spectral Densities with Application to EEG Data

• Activity Number: 427
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Chenyi Pan* and Dan Spitzner
• Companies: and University of Virginia
• Keywords: Bayesian testing ; spectral density ; kernel smoothing ; rate of testing ; EEG

Abstract

In this article, novel Bayesian testing procedures based tapering are proposed for accessing equal spectral densities of two independent stationary time series. The proposed tapering test is based on the model of log-spectral density estimates, which are calculated through kernel smoothing of log-periodogram. Bandwidth selection is explored based on the "rates of testing" criteria, a framework for measuring a test's asymptotic performance under smoothness constraints. The procedure is demonstrated on an electroencephalography (EEG) dataset of 256 channel recordings collected during the performance of sequential cognitive tasks.