對於買賣標的物在特定期間內所負起的品質擔保責任稱為保固期 (warranty period)。在市場競爭的環境下，製造商為了提高銷售量，經常訂定產品保固期來吸引消費者。然而若所訂定的保固期過長，則有可能會增加生產製造成本; 因此，如何精準地預估產品在保固期內的退貨率是製造商必需深入研究之課題。
本研究工作以資通訊產品的實驗室品質測試資料來預測其保固期內之退貨率。文獻上實驗室測試資料大多為多種產品的通過/未通過 (go/ no go) 之屬性測試資料 (attribute data)，而其退貨資料則屬於連續型壽命資料，此時若直接建構不同類型資料之預測關係式將不容易解釋其真正涵意。為了克服此困難，本研究首先將實驗室逐步應力測試之屬性資料，利用羅吉斯模型 (logistic model) 轉換成連續型資料後進行分析，文中採用empirical Bayes 層狀結構方法來結合所有產品的失效訊息以獲得模型中未知參數較精確的預測值。 最後由實例可發現本研究所提的方法比現存的文獻方法對產品退貨率有較佳的預測效果。

Quality-assurance liability of manufacture or sale matter for a specified period of time is known as the warranty period. In the market competition environment, manufacturers often adopt a warranty policy to attract their potential consumers. It is likely to increase the manufacturing and operation costs substantially if the return rate under a warranty period cannot be predicted precisely. Therefore, “how to predict the product’s return rate within the warranty period precisely” is a challenging issue to the manufacturers. The goal of thesis is mainly on the return rate prediction of information and communication technology (ICT) products, where the laboratory quality testing data are used to predict the product’s return rate under the warranty period. Most of existing literature, the step-stress laboratory testing data are recorded by a discrete-type “go/no go” pattern, while the return rate is collected by a continuous-type measurement. Hence, the physical meanings between laboratory and field data are not easy to explain due to two different sources of data. To overcome this difficulty, this thesis introduces a logistic model to transform attribute-type “go/no go” data into a continuous-type data. In addition, a hierarchical empirical Bayes procedure is also adopted to provide a better utility of all product’s information. Finally, from a real case study, it demonstrates that the proposed procedure has a better prediction performance, in comparing with the conventional approach.

[1] 林逸樵，「品質測試資料應用於保固產品之退貨率預測」，國立清華大學統計研究所，碩士論文，中華民國一零二年。
[1] Arnold, S. F. (1990). “Mathematical Statistics,” Prentice Hall, Inc., New Jersey.
[2] Hong, Y., & Meeker, W. Q. (2010). “Field-failure and warranty prediction based on auxiliary use-rate information,” Technometrics, 52, 148-159.
[3] 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.
[4] Meeker, W. Q. (1987). “Limited failure population life tests: application to integrated circuit reliability,” Technometrics, 29, 51-65.
[5] Meeker, W. Q., Escobar, L. A., & Hong, Y. (2009). “Using accelerated life tests results to predict product field reliability,” Technometrics, 51, 146-161.
[6] Suzuki, K. (1985). “Estimation of lifetime parameters from incomplete field data,” Technometrics, 27, 263-271.
[7] Tseng, S. T., Hsu, N. J., & Lin, Y. C. (2016). “Joint modeling of laboratory and field
data with application to warranty prediction for highly reliable products,” IIE Transactions, 48, 710-719.