Binary and Hybrid Response Data in Sensitivity Testing: Sequential and Bayesian Optimal Designs
Abstract
Sensitivity testing is a commonly encountered problem in defense and heavy industries. A main purpose is to estimate the critical stimulus level of an experimental object by subjecting the test specimens under a variety of stress levels. The sample size in such testing tends to be small due to high cost and testing time. Thus efficient sequential procedure for choosing stress levels adaptively is desired because it can reduce the required sample size. Due to its practical importance, this problem has received a lot of attention in the past decades. However, much less is known or has been done when the response takes a hybrid form, which has two parts. In addition to the binary response, if it is a response (e.g., explosion), there is a further measurement on the nature of the response (e.g., explosive energy). This measurement is on a continuous scale and forms the second part of the response. It can provide valuable information beyond the binary response. A major part of this project is to develop efficient sequential design procedures for conducting sensitivity testing with hybrid response. The second part of the project concerns fixed sample and sequential designs of sensitivity testing with binary data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 12, 2017
- Accession Number
- AD1058589
Entities
People
- Chien-fu J. Wu
- Roshan J. Vengazhiyil
Organizations
- Georgia Tech Research Corporation