Most programs for performing discriminant analysis provide a summary table of hits and misses in predicting group membership by using the discriminant function. The interpretation of such tables can be enhanced greatly by computing Cohen's kappa, κ, the chance corrected percentage of agreement between actual and predicted group membership. The standard error of kappa can be used to set confidence limits for the accuracy of the discriminant prediction and to test the difference in predictive accuracy for two independent samples. This was demonstrated in this article, using data previously published in a more preliminary form.
Get full access to this article
View all access options for this article.
References
1.
Barr, A. J., Goodnight, J. H., Helwig, J. H., Perkins, C. G., Sall, J. P., and Service, J.A User's Guide to the Statistical Analysis System. Raleigh, North Carolina: North Carolina State University Institute of Statistics, 1975.
2.
Cohen, J.Weighted kappa: Nominal scale agreement with provision for scaled disagreement or partial credit. Psychological Bulletin, 1968, 70, 213-219.
3.
Fenster, C. A., Wiedemann, C. F., and Locke, B.Police personality— social science folklore and psychological measurement . In Sales, B. D. (Ed.), Psychology in the Legal Process. New York: Spectrum, 1977.
