Evaluating the performance of degrees of freedom estimation in asymmetric GARCH models with t-student innovations

This work investigates the effects of using independent Jeffreys prior for degrees freedom parameter a t-student model in asymmetric generalised autoregressive conditional heteroskedasticity (GARCH) model. To capture asymmetry reaction to past shocks, smooth transition models are assumed variance. We adopt fully Bayesian approach inference, prediction and selection discuss problems related estimation Student-t propose solution based on priors which correct likelihood function. A simulated study is presented investigate how parameters GARCH affected by small sample sizes, distributions misspecification regarding sampling distribution. An application Dow Jones stock market data illustrates usefulness with errors.