Parameter Estimation Of Heteroscedastic Regression Model Based On Martingale Difference Divergence

In the general linear model or regression model,it is usually assumed that the variance is homogeneous,but in practical research,such as econometric model,the assumption is often not tenable,and for some studies,heteroscedasticity analysis is very important to understand the factors affecting heteroscedasticity,which shows that the parameter estimation of heteroscedasticity regression model has important research value.In this paper,based on the characteristic that martingale difference divergence can measure the independence of conditional mean,two constrained optimization problems are constructed,and parameter estimation is realized by solving constrained optimization problems,The consistency and asymptotic normality of parameter estimation are proved.Traditional estimation methods may have different requirements for models or variables.Simulation experiments show that our estimation method still has good estimation effect and strong robustness when the specific model form is unknown or the distribution of variables is complex.Finally,we apply the estimation method to a group of real data and analyze the actual situation...