2009-08-012024-05-17https://scholars.lib.ntu.edu.tw/handle/123456789/679552摘要：此研究計畫，將以 copula function 為基礎，建立一個能夠處理樣本選擇問題 的隨機邊界模型。與文獻中既有的模型相較，此模型的優點在於：（1）選擇 誤差可以和主模型的誤差項相關，而這也是在實證上較為可信的設定；（2） 此模型可以被現有的最大概似估計法估計；（3）此模型可以容許主方程式 的誤差項，採用多種不同的分配假設，如半常態分配、指數分配、gamma 分配 等等。我們以一個有效工資率的模型為例，說明此模型的應用。<br> Abstract: Alternative to Greene’s approach, we suggest using a copula function to model the dependency or the possible correlation between the selection equation and the SF model. The advantages of the copula based approach include (1) the selection error is allowed to be correlated with the composed error of the SF model (i.e., vi &#8722; ui) which we believe is the more plausible case; (2) it is easy to construct the likelihood function of the model to which a typical ML procedure can be applied to obtain ML estimates; and (3) the approach is easy to accommodate SF models with different distribution assumptions, such as normal-half normal, normal-truncated normal, normal-exponential, and normal-gamma distributions. We demonstrate the use of the model by estimating an efficiency wage model with sample selections using U.S. data.隨機邊界模型樣本選擇保留工資最大概似估計Stochastic frontier modelSample SelectionReservation wageCopulamaximum likelihood estimation