Linear disriminant analysis for interval data (Q880911)

The authors consider three approaches to the generalization of linear discriminant analysis (LDA) methods to multivariate interval data. In the first approach, some generalizations of the notions of ``linear combination'' and ``covariance matrix'' are made to cover multivariate intervals and LDA based on these generalizations. In the second approach, each vertex of a multivariate interval is considered as a single observation. In the third approach LDA is applied to the mid-points and ranges of the intervals. An LDA for a ``car dataset'' is considered as an example.