ERIC Number: ED365686
Record Type: RIE
Publication Date: 1993-Nov
Reference Count: N/A
Quadratic versus Linear Rules in Predictive Discriminant Analysis.
Either linear or quadratic rules may be used to derive classification equations in discriminant analysis for the purpose of predicting group membership. Generally, the decision about which rule to use is governed by the degree to which the separate group covariance matrices are unequal. An example is presented that supports the superior internal classification hit rate of quadratic rules under conditions in which the sample matrices are unequal. The superiority of quadratic internal classification results provided by SAS relative to those provided by SPSS-X is also demonstrated. Finally, it is suggested that the potential external generalizability of the classification results also must be considered when deciding whether to use linear or quadratic rules to derive classification functions. Four tables. (Contains 16 references.) (Author)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Linear Models; Predictive Analysis; Quadratic Equations; Statistical Analysis System; Statistical Package for the Social Sciences
Note: Paper presented at the Annual Meeting of the Mid-South Educational Research Association (22nd, New Orleans, LA, November 11-12, 1993).