Extreme value models for multi-variate data and its application to risk management

我们研究了多变量数据极值协同运动的建模和分析的统计方法，以及它们在极端事件影响的度量和金融及相关领域的风险管理中的应用。结果表明，状态转换模型能够灵活地根据潜在变量的状态变化将多变量数据的结构变化表示为误差协方差，在理论和实践上都是非常有用的。在金融市场的一些应用中，制度转换模型识别高风险和低风险制度，并根据当前制度和未来制度预测提出适当的投资机会和风险管理。

We investigated statistical methods for modeling and analyzing extremal co-movements of multivariate data and their applications to the measurement of the imapact of extremal events and risk management in fiance and related fields. The results indicate that the regime switching model is very useful both from theoretical and practical view points for it flexibly represents such structural changes of multivariate data as error covariances according to state changes of a latent variable. In some applications to financial markets, regime switching models identify high risk and low risk regimes and suggest appropriate investment opportunities and risk management depending on the current regime as well as future regime forecast.