Document Actions
Seminar: Bayesian Sampling Weights: Toward a Practical Implementation of the Polya Posterior
Jeremy W. Strief, Ph.D. Candidate, School of Statistics, University of Minnesota
| What | MPC Seminar Series |
|---|---|
| When |
August 27, 2007 12:15 PM
August 27, 2007 01:15 PM
August 27, 2007 from 12:15 pm to 01:15 pm |
| Where | MPC Seminar Room |
| Contact Email | mpc@umn.edu |
| Contact Phone | 612-624-8806 |
| Add event to calendar |
vCal
iCal
|
ABSTRACT: Common statistical problems encountered by Minnesota Population Center (MPC) data users are population mean estimation and regression coefficient estimation. In performing such estimation, researchers seek to apply statistical methods with good Frequentist properties, but the methods must also be practical. The Polya Posterior is a Bayesian mode of survey analysis which has good Frequentist properties. Its use, however, requires knowledge of linear algebra, Markov Chain Monte Carlo, and other advanced programming skills. In order to make the Polya Posterior accessible to MPC users, various approximations to the Polya Posterior are introduced. One approximation is based upon a new concept called Bayesian sampling weights. Not only are such weights practical to use, but, when estimating the population mean in the presence of auxiliary information, the weights can have better Frequentist properties than the Polya Posterior. Such behavior is illustrated with simulations conducted upon MPC data. Theoretical support for the Bayesian sampling weights is provided through an admissibility argument.
The National Children's Study: Design and measurement issues