ERIC Number: ED368787
Record Type: RIE
Publication Date: 1994-Jan
Reference Count: N/A
A Comparison of Best Model Selection Criteria in Multiple Regression.
Schumacker, Randall E.
The all-possible subset approach is recommended as an alternative over stepwise methods for selecting the best set of predictor variables for multiple regression. Several criteria are available for selecting the best subset model. These are compared with the principal component regression (PCR) method to investigate their usefulness for subset model selection. An applied example illustrates the comparisons. Subjects were 80 females and 76 males who participated in an early college admission program for gifted students entering the Texas Academy of Mathematics and Science. Their scores on the Learning and Study Strategies Inventory (LASSI) were analyzed; the research question was whether LASSI subscales could predict a student's college grade point average. The best subset model for each subset size with the corresponding selection criteria is presented in a table. Six tables and one figure present analysis results. An appendix contains the Statistical Analysis System program for the analysis. (Contains 22 references.) (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: All Possible Subset Approach; Principal Components Analysis; Stepwise Regression; Subset Analysis
Note: Paper presented at the Annual Meeting of the Southwest Educational Research Association (San Antonio, TX, January 27-29, 1994).