View Proposal #147
If this proposal belongs to you, you are authorized to update it. Use the menu on the right.
ID | 147 |
---|---|
First Name | Scott |
Last Name | Wood |
Institution | University of Iowa |
Speaker Category | graduate student |
Title of Talk | Model Fitting and Selection for County-Level Depression Hospitalization Rates Using Bayesian Statistical Methods |
Abstract | Researchers in the health sciences are interested in identifying and modeling the risk factors that are associated with high rates of hospitalization for depression. Being able to identify U.S. counties with high standardized hospitalization rates (SHR) would be useful in allocating federal resources. This project analyzes and critiques three potential Bayesian statistical models that can be implemented using WinBUGS software. Ordinary least squares, Poisson regression, and Bayesian conditional autoregressive (CAR) models are considered in detail. Though each has its advantages and disadvantages, qualitative and quantitative evidence suggest that the Bayesian CAR model is the optimal choice for this data. While a Bayesian CAR model will be shown to account for spatial autocorrelation and Poisson response variables, it was not as reliable as hoped for making accurate predictions at the county level. |
Subject area(s) | Bayesian statistics, spatial statistics, medical geography |
Suitable for undergraduates? | Yes |
Day Preference | |
Computer Needed? | Y |
Bringing a laptop? | N |
Overhead Needed? | Y |
Software requests | Microsoft PowerPoint |
Special Needs | None |
Date Submitted | 2/27/2006 |
Year | 2006 |