Some Practical Aspects of Exploratory Projection Pursuit

This paper shows how to choose the number of terms in a projection pursuit index and how to select a projection pursuit algorithm. It also demonstrates that Friedman’s index [J. Amer. Statist. Assoc., 82 (1987), pp. 249–266] works better than Hall’s [Ann. Statist. 17 (1989), pp. 589–605] in detecting separations or clusters.