Discriminant Analysis Via Statistical Packages

Abstract

The two aspects of discriminant analysis are briefly reviewed: predictive discriminant analysis (PDA) and descriptive discriminant analysis (DDA). Use of three popular statistical packages (BMDP, SAS, and SPSS) to obtain computational results for a PDA and for a DDA is generally reviewed. Results yielded by two BMDP procedures (7M and SM) are discussed, as are four SAS procedures (DISCRIM, STEPDISC, CANDISC, and GLM), and two SPSS procedures (DISCRIMINANT and MANOVA). It is pointed out which procedures are used to obtain PDAresults and which are used for PDAresults. Examples of printout information are given. Some of these examples pertain to misleading information-one to incorrect information, and some, of course, to very useful information. The need for nonpackage analyses to generate specific PDA and DDA information is also mentioned.

Get full access to this article

View all access options for this article.

References

Huberty, C. J (1989). Problems with stepwise methods: Better alternatives. In B. Thompson (Ed.), Advances in social science methodology (Vol. 1, pp. 43-70). Greenwich, CT:JAI.

Huberty, C. J (1994). Applied discriminant analysis. New York: John Wiley.

Huberty, C. J. & Julian, M. W. (1995). An ad hoc analysis strategy with missing data. Journal of Experimental Education, 63, 333-342.

Khattree, R. , & Naik, D. N. (1995). Applied multivariate statistics with SAS software. Cary, NC: SAS Institute Inc.

Norusis, M. J. (1993a). SPSS for windows: Advanced statistics, release 6.0. Chicago: SPSS Inc.

Norusis, M. J. (1993b). SPSS for windows: Professional statistics, release 60. Chicago: SPSS Inc.

SPSS Inc. (1993). SPSS base system syntax reference guide, release 60. Chicago: Author.

Thompson, B. (1995). Stepwise regression and stepwise discriminant analysis need not apply here: A guidelines editorial. Educational and Psychological Measurement, 55, 525-534.

White, A. P. (1994). Cross-validation of nearest-neighbor discriminant analysis-A warning to SAS users. Journal of Statistical Computation and Simulation, 49, 129-140.