The Selection And Application Of The Threshold Of POT Model

In recent years, some financial crisis and the volatility in the world’s financialmarkets occur frequently, so that the majority of financial regulators and investors infinancial asset values plummeted become particularly sensitive. The assumptionof traditional distribution is severely questioned by spikes and fat tails of thefinancial asset return series Including block sample maxima model (BMM) andthreshold model (POT), extreme value theory is the model technology ofstochastic processes of extreme value distribution and characteristics, it has theability to transcend the sample data, and can accurately the description of thedistribution of the median tail.The first chapter introduces the background of extreme value theory generated.The second chapter introduces the principle of BMM model, including the definitionof generalized extreme value function and Fisher Tippett theorem, and thendescribes in detail the structure of POT model, risk measure VaR and ESparameters estimation of generalized Pareto function. The third chapter describesseveral methods POT model threshold, such as Hill diagram method, exponentialregression model method, and nuclear goodness of fit statistic method. The fourthchapter describes the fit residuals method based on an average over function methodis proposed to improve the stability threshold. Finally, we were with the ShanghaiComposite Index January1,2006to January1,2007and the monthly price of goldin1968-2009to test the POT model described herein threshold selection method isfeasible.