k-means and Model-Based Clustering

This chapter serves as a very brief overview of classical non-robust methods for sample reduction. We will briefly review the classical k-means procedure, which can be formulated in a non-inferential framework and yields spherical clusters. We will then note that the classical k-means procedure can be cast in an inferential framework and generalized to yield elliptical clusters. The chapter will end with a discussion on choosing the number of clusters.