Development of Statistical Discriminant Mathematical Programming Model Via Resampling Estimation Techniques

This paper uses resampling estimation techniques to develop a statistical mathematical programming model for discriminant analysis problems. Deleted-d jackknife, deleted-d bootstrap, and bootstrap procedures are used to identify statistical significant parameter estimates for a discriminant mathematical programming (MP) model. The results of this paper indicate that the resampling approach is a viable model selection technique. Furthermore, estimating the MP models via resampling techniques can also improve the classification performance compared to a deterministic discriminant MP model. In this study, the deleted-d jackknife procedure was the most promising among the resampling estimation techniques examined. Copyright 1997, Oxford University Press.