Online Program

Variance Estimation of the NPMLE of the Mean with Current Status Data

*Zhen Han, University of Illinois at Chicago 
Jong Sung Kim, Portland State University 

Keywords: singly censored current status data, kernel density estimation, survival analysis, variance estimation

In bio-statistical applications, interest often focuses on the estimation of the distribution of time T between two consecutive events. In some cases, the initial event time is observed and the subsequent event is only known to happen before or after an observed monitoring time Y. This data conform to the well understood singly censored current status data, also known as interval-censored data, case 1. Currently, there are some statistical software packages such as survival, interval, and Icens that are designed to analyze randomly censored data, including this singly censored current status data. However, they either fail to directly provide an estimated variance of the estimator of the mean or produce an excessively underestimated value. In this paper, we review Huang and Wellner (1995) and propose a kernel density estimate of their variance expression of the estimator of the mean. The efficiency of our method is compared with those of bootstrap variance estimate, an open source statistical software R's survfit's variance estimate, and sampling variability. A detailed analysis of a real data illustrates the whole procedure and the R code is available upon request.