Investigating the Specification of the Distributional Assumption of the Innovations of Generalised Autoregressive Score Model with its Variants

摘要

This paper investigates the specification of the distributional assumption of the innovations of Generalised Autoregressive Score model with its variants through Monte Carlo simulation experiment with extension to international crude oil prices using three levels of volatility persistence by varying the parameters of the models. The estimates of Akaike Information Criterion and Excess Kurtosis were used to access fitness performance. The results obtained showed that when the distributional assumption of the innovations of Generalised Autoregressive Score model with its variants is either Student-t or Skewed Student-t, specifying it as Skewed Student-t distribution fits better than the other distributions considered.