本文旨在研究危險率、存活率及存活時間的計量模型，此種計量模型可 利用在各種期間資料的研究上，其重要價值。本文將之歸納為三種:參數模型、 半參數模型 (危險比例模型)、無參數模型。由於參數模型所得結果最佳，因此， 從事實證分析時，建議應先設定參數模型方法，然而一般對期間資料之分配型態 未能確知，必須加以檢定選擇，但若檢定失敗，則可進一步考慮半參數模型及無 參數模型。本文利用台灣電力及電子機械器材製造業進行上述計量模型之實證研 究，發現廠商的存活時間的分配為log-logistIc模型最佳，亦即廠商的退出的機率 模型為一例U字型，危險比例模型不適合。

　　　　　The purpose of this paper is to construct econometric models forhazard rate, survival rate and survival time. These econometricmodels are important and useful for analyzing duration data. Thepaper suggests that a parametric model shoned be established first.However, due to the shortage of information about distributionalshape, we must be careful to test whether the chosen distributionis accepted or not. Otherwise, we apply semiparametric modelinstead of parametric model. The graphical test for distributionalassumption is developed in the paper and successfully helps to selectthe parametric distribution. The econometric models are testedusing Taiwan Electronic and Electric Appliance Industry data. Theempirical results show that the survival time of firm follows loglogistic model, and the hazard rate is inverted U-shape.