An iterated function system (IFS) on the space of distribution functions is built with the aim of proposing a new class of distribution function estimators. One IFS estimator and its asymptotic properties are studied in detail. We also propose a density estimator derived from the IFS distribution function estimator by using Fourieranalysis. Relative efficiencies of both estimators, for small and moderate sample sizes, are presented via Monte Carlo analysis.
展开▼
