Advances and Applications in Statistics
Volume 53, Issue 6, Pages 625 - 645
(December 2018) http://dx.doi.org/10.17654/AS053060625 |
|
SELECTION OF VARIABLES IN MULTI-GROUP DISCRIMINANT ANALYSIS USING MATHEMATICAL PROGRAMMING AND CLASSICAL APPROACHES
Hanaa Abd El Reheem Salem and Mona N. Abdel Bary
|
Abstract: There are many variable selection procedures in traditional discriminant analysis (DA) and multinomial logistic regression (MLR). These rules enable a researcher to distinguish significant variables from non-significant ones and thus give a model classified depended solely on significant variables. Prominent among such procedures are the forward and backward stepwise variable selection criteria. No such criterion currently exists for mathematical programming (MP) techniques in discriminant analysis. The objective of the study is to select the variables in multi-group discriminant analysis using mathematical programming and stepwise discriminate analysis and multinomial logistic regression. Also, the comparison between the classical and mathematical programming approaches for variable selection in discriminant analysis, real data that were collected and reported as a practical application in (Johnson and Wichern [13]) will be used. The comparison shows that the MP model is the best. |
Keywords and phrases: selection of variables, stepwise discriminant analysis, mathematical programming, multinomial logistic regression.
|
|
Number of Downloads: 381 | Number of Views: 3478 |
|