How to Improve the Nonparametric Density Estimate in S-PLUS

Some properties of nonparametric estimates of probability density used in the interactive statistical data analysis package S-PLUS are discussed. A simple and efficient (from computational point of view) method for constructing nonparametric density estimates is suggested. In many situations, this method (based on splines) substantially dominates kernel estimates and, in particular, the estimate used in S-PLUS.