Choice of optimal smoothing parameter for nonparametric estimation of regression reliability model

The problem of nonparametric estimation of regression reliability model is considered. We consider nonparametric estimates, suggested by Beran. The main factor influencing the quality of estimates is the choice of smoothing parameter. On the example of samples, simulated from the accelerated failure time model it has been shown that the choice of smoothing parameter should be based on the difference between reliability functions corresponding to different values of the covariate, whereas the influence of the sample size and plan of experiment is not significant in the choice of smoothing parameter. In this paper we propose the algorithm of the choice of optimal smoothing parameter for non-parametric Beran estimate of regression reliability model. The algorithm is based on the minimization of standard deviation of lifetimes from nonparametric estimate of the inverse reliability function. In all considered examples the Beran estimates, obtained with the optimal smoothing parameter, turn out to be more accurate than in the case of using fixed parameter.

Keywords

Authors

References