本文試圖在不同的生產邊界估計方法及函數設定下，衡量臺灣稻作及果 樹等類作物農場之生產邊界函數及技術效率，並據以探討估計方法及函數型式之 不同對農場技術效率之影響。實證上，我們採用線性規劃法、修正最小平方法及 隨機性生產邊界法，函數型式設定則包括Cobb-DouglaS及Translog函數。實證結果 顯示:稻作農家之生產適合於具有固定模報酬之Cobb-Douglas函數。果樹農家則以 其有變動規模報酬性的Translog函數為佳。估計方法不同對技效率值影響之結果 亦顯示:同一類型農家利用不同估計方法所得到的技效率值彼此間有顯著的差 異，唯其在效率值大小乏排序性上郤相當一致。

　　　　　This paper attempts to measure production frontier function and individual farm technical efficiency for rice and fruit in Taiwan by using alternative frontier methodologies with different functional specifications. Three frontier methodologies (Probablistic Linear Programming, Corrected OLS and Stochastic Frontier models) and two functional forms (Cobb Douglas and Translog) were employed, the efficiency estimates derivedfrom these alternative models were also compared to evaluate the sensitivity of choice of the frontier models. The empirical results have indicated that the rice farm with constant return to scale was better estimated by the Cobb-Douglas form, whereas the fruit farm with decreasing return to scale was better estimated by the Translog form. The efficiency estimates from various methodologies were significantly different, neverthless the comparative renkings of individual farm efficiencies were quite consistentacross the models.