以往最常以標準常態分配建立參數模型估計金融資產報酬,而跳躍型態也比較適用於這些觀察資料的統計特徵。本研究採用道瓊工業股價指數的日報酬資料為主要之研究對象,採用較為一般化的Skewed t分配,而報酬資料的跳躍型態依然在本研究的重點之列,由於價格在短期間存在跳躍(Jump)的現象,因市場突發性重大經濟事件之發生造成資產報酬率隨機跳躍,跳躍大小會影響報酬率。因此利用ARJI模型捕捉此不連續的行為,進而本研究建立ARJI-Skewed t 模型延伸了Chan and Maheu(2002)的ARJI模型,實證結果發現: 一、透過ARJI-N及ARJI-Skewed t模型的建立,其表現能夠有5%的概似比例檢定顯著水準優於常數跳躍模型。 二、無論常數跳躍模型或ARJI模型採用服從偏態 t 分配比服從常態分配會有較高的概似比例値,惟採用服從偏態 t 分配的模型,其績效表現雖然能夠比以往採用服從常態分配的模型為佳,但是並沒有顯著性的差異。 In most financial asset returns, the most commonly used parametric specification for the return distribution is standard normal distribution, but discrete jumps in returns are better for matching some statistical features observed in the data. In this study, we use the daily return from the DJIA as research subject, and a more generalized return distribution, Skewed t-distribution, which includes the normal distribution as a special case, as the research approach. In addition the jumps in return are also included in this study. The presence of jumps exists in stock price, and the jump intensity may affect returns. Therefore this study proposes the concept of ARJI-Skewed t model, which extents the ARJI models in Chan and Maheu (2002). The findings of this study are as follows: first, the ARJI-N and ARJI-Skewed t models have better Likelihood Ratio performance than the GARCH-CJ-N and GARCH-CJ-Skewed t models. Second, the ARJI-Skewed t model’s Likelihood Ratio does not significantly higher than that of the ARJI-N model.