Spline Estimation of Conditional Variance Models with Dependent Innovations

Modeling heteroscedasticity is interesting. In this paper, we consider conditional variance models where the two sequences of explanatory variable and innovation are both strongly mixing dependent sequences. But the two sequences are mutually independent.. Under mild conditions, optimal global rate of convergence of polynomial spline estimator of nonparametric conditional variance function is obtained. Monte Carlo simulations illustrate the feasibility of the method.