Online Program Home
My Program

Abstract Details

Activity Number: 221 - Analysis of Big Dynamically Dependent Data
Type: Invited
Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
Sponsor: Business and Economic Statistics Section
Abstract #326677
Title: Dynamic Shrinkage Processes
Author(s): David Matteson* and Daniel R Kowal and David Ruppert
Companies: Cornell University and Rice University and Cornell University
Keywords: time series; trend filtering; dynamic linear model; stochastic volatility; asset pricing

We propose a novel class of dynamic shrinkage processes for Bayesian time series and regression analysis. Building upon the global-local framework of prior construction, in which continuous scale mixtures of Gaussian distributions are employed for both desirable shrinkage properties and computational tractability, we allow the local scale parameters to depend on the history of the shrinkage process. The processes inherits the desirable shrinkage behavior of popular global-local priors, such as the horseshoe prior, but provides additional localized adaptivity, which is important for modeling time series data or regression functions with local features. Efficient Gibbs sampling algorithms are employed via techniques from stochastic volatility modeling and by deriving a Polya-Gamma scale mixture representation of the process. Dynamic shrinkage processes produce superior Bayesian trend filtering estimates and posterior intervals for irregular curve-fitting of minute-by-minute Twitter CPU usage data. We develop an adaptive time-varying parameter regression model to assess the efficacy of the Fama-French 5-factor asset pricing model with momentum added as a 6th factor.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program