|
Activity Number:
|
313
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #304279 |
|
Title:
|
Prediction-Based Model Selection
|
|
Author(s):
|
Adam L. Pintar*+ and Christine Anderson-Cook and Huaiqing Wu
|
|
Companies:
|
Iowa State University and Los Alamos National Laboratory and Iowa State University
|
|
Address:
|
, , ,
|
|
Keywords:
|
|
|
Abstract:
|
Several algorithms are commonly used in regression model selection. For maximum likelihood estimation, an algorithm based on AIC or BIC may be constructed; in the Bayesian setting, Stochastic Search Variable Selection may be utilized. The current algorithms have several commonalities. The subset of regressors selected is a function of only the observed covariate values, and the goodness of the model for each subset of covariates is characterized by a single number. We introduce a variable-selection method aimed at predicting an expected response well over a user-specified range of covariate points. The proposed algorithm compares models numerically and graphically by examining the distributions of mean squared errors of prediction. An example illustrates the steps of the method and shows the importance of using a method that matches the goal of the study.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
