加速衰變試驗 (accelerated degradation test, ADT) 是工業界最常被用來推估高可靠度產品的可靠度資訊 (如產品壽命第 p 百分位數) 之重要分析工具，因此如何規劃有效率的加速衰變試驗，尤其是決定加速應力所需之最適樣本數配置，是可靠度工程師經常面臨之重要決策問題。目前文獻上有關此問題的研究方法，大多僅針對特定衰變模型 (如 Wiener, Gamma, Inverse Gaussian 衰變模型等) 進行分析。然而，對於一般化的衰變模型，缺乏有系統的探討其樣本數配置問題。為克服前述困難，本論文首先建構指數分散 (exponential dispersion, ED) 衰變模型，其優點是前述之衰變模型皆為其特例。其次，針對兩個應力水準的 ADT 問題，本論文分別在 V-optimality, D-optimality, 及 A-optimality 準則下，推導出最適 (樣本數) 配置比例之解析解。具體而言，在 V-optimality 及 A-optimality 準則下，最適配置比例皆為 ED 模型中未知參數针 及應力水準之函數；然而，在D-optimality 準則下，最適配置為高低應力都分配相同樣本數，亦即它與未知參數及應力水準皆無關。此外，本文亦深入探討此三種準則下最適樣本配置方法之相對效率 (relative efficiency)，以碳薄膜電阻器的第50 百分位數之估計值為例，採用 D-optimal 及Ａ-optimal 配置比例相對於Ｖ-optimal 配置比例之相對效率約
分別為85%及83%。

Accelerated Degradation tests (ADTs) are widely used to assess the lifetime information of highly reliable products possessing quality characteristics that both degrade over time and can be related to reliability. Hence, how to design an efficient ADT plan for assessing product’s lifetime information at normal-use stress (especially for the optimal sample-size allocation to higher test-stress levels) turns out to be a challenging issue for reliability analysts. In the literature, several papers had addressed this decision problem. However, the results are only based on a specific degradation model (such as Wiener, Gamma, inverse Gaussian models, etc.) and it lacks of a uniform approach for a general degradation model. To overcome this difficulty, we first propose an exponential dispersion (ED) degradation model which covers all mentioned-above degradation models. Next, by using V-optimality, D-optimality, and A-optimality criterion, we derive the analytical expression of the optimal sample-size allocation for a 2-stress ADT when the underlying degradation model follows an ED degradation model. The results demonstrate that the V-optimal and A-optimal allocations are the functions of unknown parameters and life-stress function, while D-optimal allocation turns out to be an equal sample-size allocation. Furthermore, we also discuss the relative efficiency of the D-optimal and A-optimal allocations with respect to V-optimal allocation and it demonstrates that the relative efficiency of D-optimal and A-optimal allocations with respect to V-optimal allocation are around 85% and 83%, respectively. Key words and phrases: Accelerated Degradation Tests; Exponential Dispersion Model; V-optimality; D-optimality; A-optimality.

[1] Abdel-Hameed, M. (1975). “A gamma wear process,” IEEE Transactions on
Reliability, 24, 152–153.
[2] Basu A. K. and Wasan, M. T. (1974). “On the first passage time process of
Brownian motion with positive drift,” Scandinavian Actuarial Journal, 1,
144–150.
[3] Chao, M. T. (1999). “Degradation analysis and related topics: some thoughts and
a review,” Proceedings of the National Science Council, 23, 555–566.
[4] Chhikara, R. S. and Folks, L. (1989). The Inverse Gaussian Distribution, Theory,
Methodology, and Applications, Marcel Dekker, Inc.
[5] Doksum, K. A. and Hόyland, A. (1992). “Model for variable-stress accelerated
life testing experiments based on Wiener processes and the inverse Gaussian
distribution,” Technometrics, 34, 74–82.
[6] Jorgensen, B. (1986). “Some properties of exponential dispersion models,”
Sandinavian Journal of Statistics, 13, 187–198.
[7] Jorgensen, B. (1987a). “Exponential dispersion models (with discussion),”
Journal of the Royal Statistical Society, Series B, 49, 127–162.
[8] Jorgensen, B. (1987b). “Small-dispersion asymptotics,” Brazilian Journal of
Probability and Statistics, 1, 59–90.
[9] Jorgensen, B. (1997). Theory of Dispersion Models, Chapman & Hall, London.
[10] Lawless, J. F. (2002). Statistical Models and Methods for Lifetime Data, John
Wiley & Sons, New York.
[11] Meeker, W. Q. and Escobar, L. A. (1998). Statistical Methods for Reliability
Data, John Wiley & Sons, New York.
[12] Nelder, J. A. and Wedderburn, R. W. M. (1972). “Generalized linear models,”
Journal of the Royal Statistical Society, Series A, 135, 370–384.
[13] Nelson, W. (1990). Accelerated Testing: Statistical Models, Test Plans, and
Data Analyses, John Wiley & Sons, New York.
[14] Ng, H. K. T., Balakrishnan, N., and Chan, P. S. (2007). “Optimal sample size
allocation for tests with multiple levels of stress with extreme value regression,”
Naval Research Logistics, 54, 237–249.
[15] Park, C. and Padgett, W. J. (2005). “Accelerated degradation models for failure
based on geometric Brownian motion and gamma process,” Lifetime Data
Analysis, 11, 511–527.
[16] Park, C. and Padgett, W. J. (2006). “Stochastic degradation models with several
accelerating variables,” IEEE Transactions on Reliability, 55, 379–390.
[17] Suzuki, K., Maki, K., and Yokogawa, S. (1993). “An analysis of degradation
data of a carbon film and properties of the estimators,” In Statistical Sciences
and Data Analysis, (Edited by K. Matusita, M. Puri and T. Hayakawa), VSP,
Utrecht Netherlands, 501–511.
[18] Tseng, S. T., Tsai, C. C., and Balakrishnan, N. (2011). “Optimal sample size
allocation for accelerated degradation test based on wiener process,” Chapter 27,
Methods and Applications of Statistics in Engineering, Quality Control, and the
Physical Sciences, John Wiley & Sons, Hoboken, New Jersey, 330–343.
[19] Tweedie, M. C. K. (1984). “An index which distinguishes some important
exponential families,” In Statistics: Applications and New Directions (Eds. J. K.
Ghosh and J. Roy), Indian Statistical Institute, Calcutta, 579–604.
[20] Wang, X. and Xu, D. (2010). “An inverse Gaussian process model for
degradation data,” Technometrics, 52, 188–197.
[21] Whitmore, G. A. and Schenkelberg, F. (1997). “Modeling accelerated
degradation data using Wiener diffusion with a time scale transformation,”
Lifetime Data Analysis, 3, 27–45.
[22] Yu, H. F. and Tseng, S. T. (1998). “On-line procedure for determining an
appropriate termination time for an accelerated degradation experiment,”
Statistica Sinica, 8, 207–220.
[23] 彭健育 (2008). “高可靠度產品之衰變試驗分析,” 國立清華大學統計學研究
所博士論文.
[24] 蔡志群 (2009). “Gamma衰變過程之設計與分析,” 國立清華大學統計學研究
所博士論文.
[25] 姚禹丞 (2010). “逆高斯與伽瑪衰變模型之誤判分析,” 國立清華大學統計研
究所碩士論文.