Using Linear versus Quadratic Rules in Predictive and Descriptive Discriminant Analysis.
Jennifer McGee · 2000
Both predictive discriminant analysis (PDA) and descriptive discriminant analysis (DDA) require a decision to pool group covariance matrices, or alternatively, to retain separate group covariance matrices when the group covariance matrices are too dissimilar to pool. Pooling the group Lovariance matrices invol-;es th3 so-called linear rule, generally preferred in predictive and descriptive analysis. Retaining separate group covariance matrices invokes the rule, resulting in a higher hit rate in PDA and a lower lambda in DDA. However, the quadratic rule is influenced by unique sampling error variance, making the generalizability of quadratic results suspect. (Contains 12 references.) (Author/SLD) Reproductions supplied by EDRS are the best that can be made from the original document. Linear vs. Quadratic Rules 1 Using Linear Versus Quadratic Rules in Predictive and Descriptive Discriminant Analysis Jennifer McGee Texas A&M University U.S. DEPARTMENT OF EDUCATION Office of Educational Research and Improvement EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) tlXis document has been reproduced as received from the person or organization originating it. Minor changes have been made to improve reproduction quality. Points of view or opiniOns stated in this document do not necessarily represent official OERI position or policy. PERMISSION TO REPRODUCE AND DISSEMINATE THIS MATERIAL HAS BEEN GRANTED BY 1 TO THE EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) Paper presented at the annual meeting of the Southwestern Psychological Association, Dallas, April 21, 2000. 2 BEST-COPYAVAILABLE Linear vs. Quadratic Rules 2 Abstract Both predictive discriminant analysis (PDA) and descriptive discriminant analysis (DDA) require a decision to pool group covariance matrices, or alternatively to retain separate group covariance matrices when the group covariance matrices are too dissimilar to pool together. Pooling the group covariance matrices invokes the so-called linear rule, generally preferred in predictive and descriptive analysis. Retaining separate group covariance matrices invokes the rule, resulting in a higher hit rate in PDA and a lower lambda in DDA. However, the quadratic rule is influenced by unique sampling error variance, therefore the generalizability of quadraticBoth predictive discriminant analysis (PDA) and descriptive discriminant analysis (DDA) require a decision to pool group covariance matrices, or alternatively to retain separate group covariance matrices when the group covariance matrices are too dissimilar to pool together. Pooling the group covariance matrices invokes the so-called linear rule, generally preferred in predictive and descriptive analysis. Retaining separate group covariance matrices invokes the rule, resulting in a higher hit rate in PDA and a lower lambda in DDA. However, the quadratic rule is influenced by unique sampling error variance, therefore the generalizability of quadratic