|
Activity Number:
|
132
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304762 |
|
Title:
|
Two-Class Prediction with Model Selection and Averaging
|
|
Author(s):
|
Wensong Wu*+ and Edsel A. Pena
|
|
Companies:
|
University of South Carolina and University of South Carolina
|
|
Address:
|
216 LeConte College , Columbia, SC, 29208,
|
|
Keywords:
|
two-class prediction ; model selection ; model averaging ; Bayesian decision function
|
|
Abstract:
|
We investigated a two-class prediction problem which is relevant in the analysis of microarray data, where of interest is to simultaneously predict the class memberships of a set of new subjects based on a set of completely observed subjects and the covariate information of the new subjects. Only a few predictor variables might be useful, and the link function relating class membership and predictor variables may not be completely known. There is therefore a need to take into account the selection of both the link function and predictor variables. A Bayesian decision function is developed using a combination of false discovery rate (FDR) and missed discovery rate (MDR) loss functions. The performance of this prediction function, which possesses a model-averaging property, is investigated and compared using a simulation study with other prediction models such as support vector machines.
|
