Optimal Design in Nonparametric Life Testing


    loading  Checking for direct PDF access through Ovid

Abstract

We consider the problem of estimating the common survival function S0 of independent objects. The successive failure times of the objects can be observed during a given time interval [0, T]. In order to get information more quickly, the observer has the freedom to shorten the life duration of the observed objects by applying a predictable stress Z = (Zt)t ∈ [0,T]. The stress intensity is governed by a certain cost constraint. We explore this estimation problem within its natural stochastic control setting.By choosing a nonparametric Bayesian setting, we can formulate simultaneously the problem of estimating S0 as → ∞ and the choice of Z. Under some restrictions, we propose an hybrid procedure Ŝ0(Z) for estimating S0, combining the Nelson–Aalen estimator and a dynamic programming approach in order to choose the stress Z in an optimal way.

    loading  Loading Related Articles