B0921
Title: Shrinkage of time-varying effects in panel data models
Authors: Helga Wagner - Johannes Kepler University (Austria) [presenting]
Abstract: Regression models for panel data with time-varying effects in a Bayesian framework are considered. We implement shrinkage of regression effects as well as the process variances of the effects to distinguish between constant and time-varying effects as well as significant and insignificant effects by appropriate priors distributions. The inference is accomplished via MCMC using ancillarity-sufficiency interweaving which leads to huge improvements in sampling efficiency. The model is applied to analyse panel data on the annual incomes of mothers returning to the job market after maternity leave.
