本文主要目的乃在運用無參數分析法來驗證需求資料的一致性，並用計量經濟 技術估測不同恩格爾函數型態，以找出最合適的需求體系模型。實證分析資料包括 四種肉類的消費支出，資料取自台灣糧食平衡表和農產品成本與物價統計，樣本期 間為民國六十年至七十八年。 無參數需求分析的結果，顯示樣本期間內四種肉類消費資料，滿足效用極大化 條件，但此資料之效用函數不具有固定之貨幣邊際效用特性，亦乏效用函數同位性 之性質。根據最大概似法，以迴複程序估測六種需求體系之恩格爾函數，實證結果 顯示二次式的估測函數值，高於一次式的函數型態，惟其值差異不大。利用概似比 值檢定法，在α＝0.01水準下，PIGLOG需求體系與二次式需求體系和PIGL並無統計 顯著性差異。因此，在考慮資料一致性，需求理論限制條件，與實證需求參數估測 簡便性下， PIGLOG需求體系模式應該是樣本期間，四種肉類需求體系函數模式的最 佳撰擇。

　　　　　The purpose of this study is to employ nonparametric procedures for examin- ing data consistency and to find an appropriate flexible demand system for model- ing consumer behavior. Data used in this study are obtained from Taiwan food balance sheet and Taiwan agricultural price and costs monthly. Sample periods from 1971 to 1989. Four meat commodities are included. The results of nonparametric testing show that the data satisfy the GARP, but violate the cyclical monotonicity and HARP. The mazimum likelihood estima- tion is used to estimate the parameters of Engel curves of demand systems. The various functional froms of Engel curves are tested against each other by means of likelihood ratio test. The empirical results reveal that quadratic functional forms of demand systems performed better than linear functional form in terms of estimated function values. The results of likelihood ratio tests alro indicate that the PIGLOG, the GQES, and the PIGL demand systems are not statistically significant at significant level α＝ 0.01. The PIGLOG demand system, therefore, is suggested to be the best functional form for our empirical data.