Abstract: In this paper,an extended generalized orthogonal GARCH Model,which can reflect the asymmetry and leverage of volatility,is proposed based on that of Vander Weide (2002) Then,a new estimation method for the multivariate GARCH models from the mutual information theoretic view point is considered.The relationship between the statistical dependence in standardized residuals and the maximum likelihood,when estimating multivariate GARCH models,is revealed Based on that,the different statistic ways are proposed in the framework of the extended generalized orthogonal GARCH models,for purpose of implementing the models based on the estimation of mutual information minimization.These metho dologies can therefore be easily applied to high dimensional systems,where likelihood based estimation will run into computational problems.According to 15 global stock returns indexes,empirical research is included to illust rate the model and the estimation method.

Key words: mutual information, multivariate GARCH models, leverage effects, parameter estimation, dynamic conditional correlation