- 學位論文
技術指標對股價呈現碎形小波轉換的影響—Linear、LSTARX與 ESTARX模型之比較
The effect of technical indicator on stock price showing fractal wavelet transforms — A comparison among Linear, LSTARX and ESTARX models
摘要
Teräsvirta, Dick and Medeiros ( 2003 )的研究指出LSTARX模型較類神經網路模型具有較佳的覆蓋機率範圍，引發本研究採用LSTARX與ESTARX模型找出影響台灣股市股價取自然對數後作碎形維度小波轉換的關鍵技術指標變數，進而驗證兩模型是否為有助於進行股價報酬率小波轉換預測的研究動機。在無堅強的理論模型基礎下要選擇最適“解釋變數-技術指標”，遂採用逐步迴歸的方式找出影響大盤、電子與金融三種不同類股股價指數報酬率的技術指標之線性迴歸模型，然後依這些變數的顯著性落遲期，應用LSTARX模型與ESTARX模型進行小波轉換研究。實證結果顯示，LSTARX與ESTARX模型較Linear模型在股價報酬率的預測值上相較於實際值有較高相關的上下振幅連動性；在LSTARX模型下三大類股的小波轉換參數均具有顯著性，而在ESTARX模型下三大類股的解釋變數落遲期與其移轉函數之變數均存在顯著性。換言之，LSTARX模型支持三大類股股價報酬率走勢均存在小波轉換，而ESTARX 模型則有助於找到控制股價報酬率呈現小波轉換的重要自變數。
並列摘要
Based on the conclusion, derived by Teräsvirta, Dick and Medeiros(2003), that LSTAR model appears to have better probability coverage than the neural network model, this paper adopts LSTARX and ESTARX models to find out the technical indicator variables influencing little wavelet transfers from stock price fractal dimension and use them to forecast little wavelet in reverse point of stock porice. Under no strong theorical model, we first use stepwise regression model to obtain statistically significant exogenous technical indicator variables affecting returns of stock price index, from TAIEX, electronics-index and finance-index. Further using these significant variables to form their lag terms and LSTARX and ESTARX models, then we can analyze wavelet in reverse point of stock price. Empirical evidences show that LSTARX and ESTARX models outperform linear model. For LSTARX model, the location parameters of little wavelet reversing are identified significant. And for ESTARX model, transition function variables and variables’ lag terms are significant. That is to say, LSTARX model supports there are little wavelet transition reversed in three major type stock price index returns and ESTARX model can find out the important independent variables to explain the little wavelet transfer of the stock price return.