Linear Hypotheses and Linear Discriminant Analysis

The content of this chapter is decision among alternatives which can be described by linear models. The first part discusses classical linear hypothesis testing and some traditional applications such as analysis of variance, while the second part points to the “modern” view of linear decision theory centering on linear separability and related algorithms such as the theory of optimal separating hyperplanes. A clever generalization of this idea will lead to nonlinear support vector machines.