A generalized quantile estimator under censoring

Abstract

Based on right-censored data from a lifetime distribution F₀ a smooth alternative to the product-limit estimator as a nonparametric quantile estimator of a population quantile is proposed; The estimator is a "generalized product-limit quantile" obtained by averaging appropriate subsample product-limit quantiles over all subsamples of a fixed size. Under the random censorship model and some conditions on Fop it is shown that the estimator is consistent and has the same asymptotic normal distribution as the product-limit quantile estimator. A small Monte Carlo simulation study shows that there exist some values of the subsample size for which the estimator performs better than the product-limit quantile estimator in the sense of estimated mean squared errors.

Keywords:

• random censoring
• product-limit quantity estimator
• generalized product-limit quantitp estimator
• median Survival time
• estimation: nonparametrA-c quantize estimation