Small-Sample Properties of Kernel Density Estimators from Randomly Right-Censored Data.
SOUTH CAROLINA UNIV COLUMBIA DEPT OF MATHEMATICS AND STATISTICS
Pagination or Media Count:
The small-sample behavior of two kernel-type density estimators which have been proposed in the literature for randomly right-censored samples is investigated via Monte Carlo simulations. The extensive simulation study was performed for five families of life distributions, two different censoring distributions, three kernel functions, and several bandwidth sequences and for sample sizes from n20 to n300. The simulation results reinforce previous theoretical results for the estimators and lead to conjectures about their general behavior asymptotically as well as for small samples. A comparison of the two density estimators is also indicated.
- Numerical Mathematics