保固期 (warranty period) 是對於產品本身的一種保障政策。在市場高度競爭的環境下，製造商為獲取消費者對其產品的信賴而訂立保固期，亦即產品在保固期內失效時，製造商必須提供消費者免費維修或更換新品之服務。因此，保固期內產品的退貨率 (warranty return rate) 便是製造商之重要研究課題。目前的研究工作大多針對單一產品，以實驗室的產品品質測試資料 (如加速壽命試驗) 來進行其保固期內之退貨率預測推論。然而，以消費性的資通訊 (ICT) 產品為例，其實驗室測試資料大多為多種產品的通過／未通過 (Go/ no go) 資料，且每個產品只有極少數測試失效數據可供使用。在此情況下，現存文獻並無適當的解決辦法。本論文以平板電腦為實例，對於實驗室多數無失效之數據，利用 “層狀結構”(Hierarchical Structure) 有效地結合所有產品的失效資訊，來進行實驗室失效模型中的參數估計，其優點是即便在實驗室無失效數據的情況下，皆能提供較傳統最大概似估計值更合理的推論，且對退貨率之預測能有較好的預測效果。

Warranty policy is a useful tool for the manufacturers to compete with others. When a product fails during the warranty period, manufacturers shall provide their consumers with free charge service for the products. Therefore, the return rate prediction during the warranty period is an important decision problem for manufacturers. Recently, several research works focus on using the in-field laboratory testing data to predict the return rate during the warranty period. However, the results are very restricted to a single-product with continuous measurements for laboratory data. For the consumptive information and communication technology (ICT) products, however, “Go/ no go” laboratory data is very common. In addition, laboratory data may only have very few failed data during the test. In this situation, there is no suitable tool for analyzing this kind of data. In this study, taking tablet computer for example, we propose a “Hierarchical Structure” to incorporate all failure information so that the parameters in the laboratory failure model can be estimated efficiently. The advantage of the proposed method is that even for zero-failure data, this method still provides us a robust estimation for the process parameters, and gives us a precise prediction on the return rate.

第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 1
1.3 研究架構 2
第二章 文獻探討與問題描述 3
2.1 文獻回顧 3
2.2 使用率模型之連結方法介紹 4
2.3 本研究之資料型態簡介 7
2.4 實例探討 9
2.5 問題描述 12
第三章 模型建構及退貨率之預測分析 15
3.1 實驗室測試資料建模 15
3.1.1 失效模型 15
3.1.2 層狀結構 18
3.1.3 實例說明 21
3.2 保固期內退貨資料建模 23
3.3 參數連結與預測方法 27
3.3.1 參數連結 27
3.3.2 預測方法 29
3.4 EBE與MLE之比較 30
3.4.1 參數連結比較 30
3.4.2 預測績效比較 34
第四章 結論與後續研究 37

[1] Agresti, A., & Hitchcock, D. B. (2005). “Bayesian inference for categorical data analysis,” Statistical Methods and Applications, 14(3), 297-330.
[2] Arnold, S. F. (1990). Mathematical Statistics, Prentice Hall, Inc., New Jersey.
[3] Hong, Y., & Meeker, W. Q. (2010). “Field-failure and warranty prediction based on auxiliary use-rate information,” Technometrics, 52, 148-159.
[4] Lawless, J. F., Crowder, M. J., & Lee, K. A. (2009). “Analysis of reliability and warranty claims in products with age and usage scales,” Technometrics, 51, 14-24.
[5] Meeker, W. Q. (1987). “Limited failure population life tests: application to integrated circuit reliability,” Technometrics, 29, 51-65.
[6] Meeker, W. Q., Escobar, L. A., & Chan, V. (2001). “Using accelerated tests to predict service life in highly variable environments,” In ACS Symposium Series (Vol. 805, pp. 396-413). Washington, DC; American Chemical Society; 1999.
[7] Meeker, W. Q., Escobar, L. A., & Hong, Y. (2009). “Using accelerated life tests results to predict product field reliability,” Technometrics, 51, 146-161.
[8] Suzuki, K. (1985). “Estimation of lifetime parameters from incomplete field data. Technometrics,” 27, 263-271.