主讲人：张新雨教授(Academy of Mathematics and Systems Science, Chinese Academy of Sciences)

时  间：12月5日（周五）下午15:30-16:30

地  点：必赢76net线路官网北一区文科楼708教室

摘  要： A Mixture copula is a linear combination of several individualcopulas.It is able to generate dependent  structures  that  do  not belong to existing copula families. This makes it useful in modelling the dependent structures in financial data,as  in  empirical  studies  different  pairs  of  markets may exhibit quite different dependent structures. Therefore, rather than selecting one optimal copula through certain criteria we propose using a model  averaging approach to estimate financial daa dependent structure in a mixture copula framework. We select weights (for averaging)through aJ-fold Cross-Validation procedure. We prove that the model average estimator  is  asymptotically  optimal  in the sense of minimizing squared estimation  loss.Simulation results show that the model averaging  approach significantly outperforms some competing methods when the working mixture model is misspecified. Using 12years’ daily returns of four developed economies’ stock indexes (United States, UnitedKingdom, Hong Kong and Japan), we show that the model  average approach givesmore  reasonable  estimations  of  their dependence structures than other competing economies.