| Summary: | 碩士 === 國立屏東科技大學 === 農企業管理系 === 92 === In recent years, the risk management becomes increasingly important. Value at Risk is the most weighted risk tool in the risk management. The general financial sequence are generally fat-tailed, thus, the model which is hypothesize by normal distribution can’t estimates the risk value accurately. The Extreme Value Theory (EVT) is shown to be able to grasp the fat tail phenomenon very well. This paper uses Peaks over Threshold Model (POT) of EVT to estimate the portfolio risk which is mixed with corn, wheat and soybean. Besides, The POT Models for both Parametric GPD distribution and Semi-parametric Hill estimator are employee. Besides, QQ-plot, ME-plot, Hill-plot and bootstrapping MSE threshold selection method are examined for the most suitable threshold selection method.
The empirical results show that two are equally well POT models perform better than the competitive models. The two of POT models at significance level 1% and 5%, for both back test and forward test. However, the GPD distribution is more stable, especially for the probability closed to the tail. As to the threshold selection methods, ME-plot outperforms the other on the average, though there’re no significant differences among the methods.
|
|---|
