Nonparametric Bayesian Particle Learning with Applications to Income Volatility

• Activity Number: 74
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Julie Novak*+ and Tung Phan and Shane Jensen
• Companies: Wharton School and Wharton School and University of Pennsylvania
• Keywords: Bayesian Nonparametric ; Income Volatility ; Dynamic Linear Modeling ; Particle Learning ; Heirarchical Dirichlet Process

Abstract

We propose a new method for estimating hidden states and parameters of state-space models. This approach combines particle learning and hierarchical Dirichlet process (HDP) to provide an efficient way to analyze hierarchical data. We compare our method to others, such as combining Kalman Filters and HDPs, and analyze model performance. We then apply our method to labor incomes from the core sample of the Panel Study of Income Dynamics (PSID) data set to demonstrate the importance of our new method.