Ghosh, Anil K.; Chaudhuri, Probal; Sengupta, Debasis (2006, ASA/ASQ) Indian Statistical Institute, Kolkata, India;
[This abstract is based on the authors' abstract.]Using a kernel density estimate involves properly selecting the bandwidth parameter. However, the bandwidth that is optimum for the mean integrated square error of a class density estimator may not always be good for discriminant analysis. Therefore, in a multiclass problem, it is more useful to look at the results for different scales of smoothing for the kernel density estimates than to concentrate on a single optimum bandwidth for each population density estimate. A multiclass approach with a graphical device leading to a more informative discriminant analysis is presented. The method allows the flexibility of using different bandwidths for different pairs of competing classes while reducing the computational burden for cross-validation-based bandwidth selection. Benchmark examples illustrate the proposed methodology.
Density estimation,Pairwise comparisons,Posterior,Probability function,Statistical weights