Accelerated Life Tests That Maximize Shannon Information,
Abstract
The current literature on accelerated life tests emphasizes issues of inference and extrapolation under a given design, rather than the optimal design of an accelerated test. In this expository paper, we outline a framework for a coherent approach for the conduct of accelerated tests. Our approach is based on the Bayesian paradigm for experimental designs under linear models and involves the specification of a utility function. For the latter we adopt Shannon Information between predictive densities. The optimal design is then selected via the principle of maximum expected utility. We illustrate the approach with some special cases. (MM)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1988
- Accession Number
- ADA293944
Entities
People
- Isabella Verdinelli
- Nick Polson
- Nozer Singpurwalla
Organizations
- George Washington University