本研究利用非線性模型分析臺灣地區香蕉、鳳梨、木瓜、蓮霧、楊桃，香瓜、洋香 瓜、及梨子的價格。在單根檢定方面，本研究利用Enders單根檢定程序，以最適落遲期數及 檢定型態對個別水果的價格及可供消費量進行完整的平穩性測定。檢定結果顯示除了香蕉的 可供消費量不具平穩性之外，其餘水果之價格及可供消費量序列均具平穩性。香蕉可供消費 量序列經過一期差分處理後亦呈現平穩。在非線性模型方面，本研究選取ESTAR及LSTAR 模型分析個別水果價格，結果顯示ESTAR模型及LSTAR模型整體配置程度佳。樣本外預測 除蓮霧及梨子的實際值與預測值差距較大之外，其餘水果品項的預測誤差約在每公斤5元以 內。RMSPE及Theil不等係數顯示模型的樣本內及樣本外三期的預測能力良好。

　　　　　Nonlinear autoregressive models are utihzed in this study to analyze the asymmetric price fluctuations of bananas, pineapples, papayas, wax apples, carambolas, suskmelons, cantaloupes, and pears in Taiwan. For the stationarity tests, proper functional forms are selected based on a procedure suggested by Enders Fruit prices and disappearances are tested for stationarity with appropriate lag lengths Results indicate that all the series are stationary besides the disappearances of bananas, which have become stationary after taking the first difference. For the nonlinear price analyses, the logistic smooth transition autoregressive model (LSTAR) and the exponential smooth transition autoregressive model (ESTAR) are applied. Overall, these two models fit the data well Out-of-sample predicted prices of wax apples and pears are relatively more deviated from respective real prices. For other fruits, forecasting errors are less than NT$5 per kilogram for the out-of-sample forecasting. The RMSPE and the Theil's inequality coefficients that the models are capable to make reliable predictions.