Simulated Analysis of Bivariate Extreme Mixed Model

Bivariate extreme mixed model can not reach the complete dependence of extreme variables, so it has a restriction in application. However, bivariate extreme mixed model still a good model for some dependence. In this paper we firstly introduce some basic knowledge about bivariate extreme mixed model. Specifically, we aim to assess the effects through a simulation study when a BEV distribution with mixed model dependence is fitted to data from other bivariate extreme value copula. As a result, if we measure dependence by Kendall’s τ, we find that to some extent mixed model can capture the dependence of other models. And for asymptotically independent model, the bias in the marginal parameters is not severe. At last, we use mixed conditional model and GEV conditional model to analysize the data about the log-daily returns of two exchange rates: UK sterling against both the US dollar and the Canadian dollar.