Premium
Reliability models and inference for systems operating in different environments
Author(s) -
Hollander Myles,
Peña Edsel A.
Publication year - 1996
Publication title -
naval research logistics (nrl)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.665
H-Index - 68
eISSN - 1520-6750
pISSN - 0894-069X
DOI - 10.1002/(sici)1520-6750(199612)43:8<1079::aid-nav3>3.0.co;2-b
Subject(s) - estimator , covariate , quantile , reliability (semiconductor) , inference , series (stratigraphy) , function (biology) , statistics , mathematics , econometrics , computer science , artificial intelligence , paleontology , power (physics) , physics , quantum mechanics , evolutionary biology , biology
We consider a class of models for the reliability function of a series system. The models incorporate dependence of the function on environmental covariates. On the basis of censored data obtained by monitoring several series systems operating under various sets of values of the covariates, estimators are derived of the reliability function of a series system operating under a different set of values of the covariates. Estimators of various functionals of the reliability function, such as the trimmed mean and quantiles, are also presented. Asymptotic properties of the estimators are obtained with the use of the framework of counting processes and martingales. The loss in efficiency in estimating the regression parameters and the reliability function is also examined when one assumes a more general semiparametric model, when in fact the true model belongs to a more restricted model. We illustrate these procedures with the use of accelerated failure‐time data. © 1996 John Wiley & Sons, Inc.