|
Activity Number:
|
290
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #305790 |
|
Title:
|
Power Comparisons for Model Search, Identification, and Discrimination
|
|
Author(s):
|
Ying Luan*+ and Subir Ghosh
|
|
Companies:
|
University of California, Riverside and University of California, Riverside
|
|
Address:
|
4997 Brookhill Place, Riverside, CA, 92507,
|
|
Keywords:
|
equivalence ; fractional factorial plans ; interaction effects ; linear model ; permutation matrix ; power
|
|
Abstract:
|
This presentation considers the problem of model search, identification and discrimination between models from a specified class of models. The models within this class have some common parameters and one distinct parameter in each model. The power for testing uncommon parameter is used as our criterion function. We present some characterizations for the equivalence of models based on the components of the non-centrality parameter for the distribution of the test statistic. The characterizations are given in terms of a vector h dependent only on the design matrix. As an illustrative example, four fractional factorial plans for a 2^5 factorial experiment are considered. The common parameters are the general mean & the main effects, and the uncommon parameters are two factor interactions one for each model.
|
