單根檢定已成為實證分析中用以檢定時間序列資料平穩性的基本檢定方法，以DF 及ADF檢定法最常被研究所採用。單根檢定在應用時需考量落遲期數及檢定型態，以減少在 判定上產生誤差的可能性。本研究以AIC落遲期數選取準則及Enders單根檢定程序對臺灣十 種水果價格進行平穩性測定，實證結果顯示十種水獰價格的最終檢定結果皆具平穩性，其中 香蕉、鳳梨、西瓜、木瓜、蓮霧、香瓜、洋香瓜、及梨子價格的最適檢定方程式型態含截距 及趨勢項，落遲期數分別為1期至5期；葡萄價格的最適檢定型態為含截距項、但不含趨勢 項的方程式型態，落遲期數為4期；楊桃價格的最適檢定型態不含截距及趨勢項，落遲期數 為6期。

　　　　　Stationarity is a certain condition to be maintained in time-series analyses. In applying the Dickey-Fuller or the Augmented Dickey-Fuller tests, appropriate functional form and optimal lag lengths need to be determined. This study utilizes a procedure suggested by Enders to test stationarity of fruit prices in Taiwan. For each functional form, Akaike Information Criterion is used to choose the proper lag length. Results indicate that for carambola prices, appropriate functional form for the stationarity test is the one without intercept or trend, and the lag length of six is appropriate. For grape prices, appropriate functional form is the one with the intercept but without trend, and four legs seem to be optimal. For banana, pineapple, watermelon, papaya, wax apple, muskmelon, cantaloupe, and pear prices, appropriate functional forms are with intercepts and the trend variables. Optimal lag lengths for these fruit prices are between one and five. The results of stationarity tests reveal that the prices of these fruits are stationary.