Pointwise Nonparametric Maximum Likelihood Estimator of Survivor Functions Under Stochastic Ordering Constraint

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: ENAR
• Author(s): Yong Seok Park*+ and Jeremy Taylor and John David Kalbfleisch
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: pointwise C-NPMLE ; survivor function ; stochastic ordering ; censored data ; constrained estimator

Abstract

We often encounter the situation of estimating survivor functions with prior knowledge that the survivor function of one population is less than that of another at all times. This type of constraint is called stochastic ordering. In this paper, We propose a new constrained nonparametric estimator - the pointwise C-NPMLE. The estmator is defined at each time t by obtaining the estimates of the survivor functions subject to constraints at time t only. Simulation studies show that this new estimator has better finite sample properties and can overcome problems that arise from the constrained nonparametric MLE and Rojo's(2004) estimator.