A Statistical Analysis of Some Models Used in Accelerated Life Tests.
Abstract
This paper considers the problem of inference for the parameters of three mdoels used in accelerated life tests. The first part obtains orthogonal least squares estimators for the parameters of the models. The second part considers a Bayesian analysis of the two parameters of the power rule model. The absolutely continuous bivariate exponential (ACBVE) distribution is assigned as a joint prior on the power rule parameters. The analysis proceeds along the lines dictated by Box and Tiao in their book, Bayesian Inference in Statistical Analysis. Thus, location parameters are introduced in the ACBVE so that the prior is shifted to a position where the likelihood is appreciable. For computational convenience, the joint prior is discretized over regions of the parameter space where the likelihood is appreciable. This approach allows some generality in the choice of joint priors. Using the approach, conclusions are reached pertaining to the robustness of the inferences with respect to assumptions about the Weibull shape parameter. (Author
Document Details
- Document Type
- Technical Report
- Publication Date
- May 08, 1977
- Accession Number
- ADA041001
Entities
People
- Henry D. Kahn
Organizations
- George Washington University