A2070
Title: A generalized Bayesian approach to multiple changepoint analysis
Authors: Michael Jauch - Florida State University (United States) [presenting]
Yuhui Wang - Florida State University (United States)
Joshua Loyal - Florida State University (United States)
Andrew Thomas - University of Iowa (United States)
Abstract: A generalized Bayesian method for multiple changepoint analysis is introduced, employing a loss function inspired by multinomial logistic regression. The method does not require specification of the data-generating process and avoids restrictive assumptions on the nature of changepoints. From the joint posterior distribution, simultaneous inference can be made on the locations of changepoints and the coefficients of a multinomial logistic regression model for distinguishing data across homogeneous segments. The multinomial logistic regression coefficients provide a familiar means of interpreting potentially complex changes. To select the number of changepoints, posterior summaries are leveraged that measure whether the multinomial logistic classifier can distinguish data from either side of a potential changepoint. To simulate from the generalized posterior distribution, a Gibbs sampler based on Polya-Gamma data augmentation is presented. The accuracy and flexibility of the method are assessed through simulation studies featuring different types of changes, and its interpretability is demonstrated through applications to financial network data and topological data derived from nanoparticle videos.
