TY - JOUR
T1 - Accelerated destructive degradation tests robust to distribution misspecification
AU - Jeng, Shuen Lin
AU - Huang, Bei Ying
AU - Meeker, William Q.
N1 - Funding Information:
Manuscript received April 30, 2010; revised December 17, 2010; accepted March 08, 2011 Date of publication July 14, 2011; date of current version December 02, 2011. This research is supported by the Grant from National Center for Theoretical Science (South) and from the University Advancement Project of National Cheng Kung University, Taiwan. Associate Editor: J. C. Lu. S.-L. Jeng is with the Department of Statistics, National Cheng Kung University, Tainan, Taiwan (e-mail: [email protected]). B.-Y. Huang is with the Cathay Life Insurance Company, Taipei, Taiwan. W. Q. Meeker is with the Department of Statistics, Iowa State University, Ames, IA, USA. Digital Object Identifier 10.1109/TR.2011.2161051
PY - 2011/12
Y1 - 2011/12
N2 - Accelerated repeated-measures degradation tests (ARMDTs) take measurements of degradation or performance on a sample of units over time. In certain products, measurements are destructive, leading to accelerated destructive degradation test (ADDT) data. For example, the test of an adhesive bond needs to break the test specimen to measure the strength of the bond. Lognormal and Weibull distributions are often used to describe the distribution of product characteristics in life and degradation tests. When the distribution is misspecified, the lifetime quantile, often of interest to the practitioner, may differ significantly between these two distributions. In this study, under a specific ADDT, we investigate the bias and variance due to distribution misspecification. We suggest robust test plans under the criteria of minimizing the approximate mean square error.
AB - Accelerated repeated-measures degradation tests (ARMDTs) take measurements of degradation or performance on a sample of units over time. In certain products, measurements are destructive, leading to accelerated destructive degradation test (ADDT) data. For example, the test of an adhesive bond needs to break the test specimen to measure the strength of the bond. Lognormal and Weibull distributions are often used to describe the distribution of product characteristics in life and degradation tests. When the distribution is misspecified, the lifetime quantile, often of interest to the practitioner, may differ significantly between these two distributions. In this study, under a specific ADDT, we investigate the bias and variance due to distribution misspecification. We suggest robust test plans under the criteria of minimizing the approximate mean square error.
UR - http://www.scopus.com/inward/record.url?scp=82455162366&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=82455162366&partnerID=8YFLogxK
U2 - 10.1109/TR.2011.2161051
DO - 10.1109/TR.2011.2161051
M3 - Article
AN - SCOPUS:82455162366
SN - 0018-9529
VL - 60
SP - 701
EP - 711
JO - IEEE Transactions on Reliability
JF - IEEE Transactions on Reliability
IS - 4
M1 - 5953546
ER -