Inference and Asymmetric GARCH-Model with a New Distributed Innovation

A novel generalized-odd-generalized exponentiated skew-t (GOGEST) innovation density for the generalized autoregressive conditional heteroskedasticity (GARCH) models is proposed. The features of the proposed distribution were derived. The parameter estimates of the proposed distribution through simulation were carried-out with maximum likelihood estimation technique. The performance of the asymmetric GARCH-GOGEST model relative to five other asymmetric GARCH-various existing innovation densities in volatility modeling was investigated using the Bitcoin log-returns. The empirical results showed that the asymmetric GARCH-GOGEST models were superior over the other asymmetric GARCH models. However, the threshold GARCH-GOGEST model outperformed the other models in terms of volatility predictability (out-of-sample).