On the Interpretation of Individual Variables in Multiple Discriminant Analysis
针对金融分类问题中多元判别分析无法假设组协方差相等的情况，提出一种基于条件删除法的变量重要性排序方法，帮助研究者评估个体变量的相对重要性。
A number of articles have recently appeared in the literature dealing with the application of multiple discriminant analysis (MDA) to classification problems in the area of finance [2, 3, 7, 8, 13, 15]. The problem of assessing the importance of individual variables was an issue in these papers. The objective of this paper is to develop a ranking procedure for assessing the relative importance of the individual variables when it cannot be assumed that the group covariance matrices are equal. It is assumed that the analysis is for two groups. The ranking procedure we suggest relies upon the conditional deletion procedure based on a statistic used for solving the multivariate Behrens–Fisher problem.