常用的廠商績效評估方法可歸納為兩種,其一為資料包絡分析法,其二為隨機前緣分析法,後者必須先宣告函數型式,再以最大概似法校估參數,並進一步估計各廠商之效率值,但以最大概似法校估之最大缺失,是所估計之參數常因符號不正確而違反生產經濟學之正規條件,導致奇怪的結論。為改正此一缺失,本研究嘗試改以貝氏隨機前緣分析法來校估參數及廠商之效率值,按貝氏之優點是得以強加限制條件於隨機前緣模型上,校估之結果符合經濟學之正規條件。在實證方面,本研究以24家歐盟國家的鐵路公司2006至2008年的營運資料,並宣告產出距離函數以估計廠商之營運績效。實證結果顯示,貝氏隨機前緣分析法確實能得到較為令人滿意之結果,根據實證分析結果,本研究提出若干結論與建議,以供後續研究之參考。
The most commonly applied commercial performance evaluation methods comprise the data envelopment analysis (DEA) method and the stochastic frontier analysis (SFA) method. Of these methods, the SFA method sequentially requires the researcher to establish functional forms, calibrate and estimate relevant parameters using the maximum likelihood (ML) method, and estimate the efficiency values of the various businesses. However, the estimation of parameters using the ML method often yields incorrect signs, which consequently violates the regularity conditions and results in convoluted conclusions. To rectify this drawback, this study endeavors to employ an alternative Bayesian SFA method to calibrate and estimate relevant parameters and efficiency values of the businesses. The advantage of Bayesian methods is that conditions are restricted to the SFA model. This facilitates the calibration and estimation results to comply with economic regularity conditions. For the empirical research, this study collected the operational data (2006 to 2008) of 24 railway companies located in countries that were members of the European Union. Subsequently, this study established output functions to estimate the operational performance of businesses. The results suggested that the proposed Bayesian SFA method successfully yielded results that were more satisfactory. This study further provided relevant conclusions and suggestions based on these results, which can be used as a reference for future research.