B1936
Title: Asymptotic normality of the deconvolution kernel density estimator based on strong mixing and right censored data
Authors: Shan Sun Mitchell - The University of Texas at Arlington (United States) [presenting]
Abstract: The challenge of estimating the distribution of $Z$ is considered, a variable that cannot be directly observed. Instead, it is only observed through the presence of an error-contaminated variable $X$, defined as $X=Z+E$. In this context, $X$ represents an observable right-censored survival time with an unknown density function, while E is a measurement error known to follow a particular distribution. Assuming a sequence of sample $X$ satisfies the strong mixing condition, a method for estimating the unknown density is proposed, combining the ideas of deconvolving kernel density estimator and inverse-probability-censoring weighted average. The asymptotic normality is also established of this estimator under two distinct assumptions: one where the tail behavior of the characteristic function of E is considered 'supersmooth,' and the other where it is considered 'ordinarily smooth'.
