Bayesian GARCH models for Nigeria Covid-19 data

Abstract

Due to the high demand for accurate volatility forecasts, there has been an immense interest amongst both practitioners and researchers to model time varying volatility series. Unlike many researches on GARCH models, this paper centres on the use of generalized autoregressive conditional heteroskedasticity (GARCH) models for medical data. With GARCH models, issues of no (conditional) heteroskedasticity, no error autocorrelation, linearity, and parameter constancy are adequately tested. Bayesian estimation method is used to estimate the parameters of GARCH models based on daily COVID-19 total cases and new cases in Nigeria, for the period (28^th February, 2020) when the index case came to Nigeria through to when the second wave (19^th January, 2021) of the COVID-19 cases in Nigeria began to rise. Taking advantage of Bayesian approach to estimating the parameters of a model, two Bayesian generalized autoregressive conditional heteroskedasticity (GARCH) models were used in capturing the volatility dynamics of the daily total cases and daily new cases of COVID-19 in Nigeria based on their conditional distributions. From analysis, diagnosis tests, results revealed that the Bayesian Student-t GARCH (1, 1) model performed better than the Bayesian Normal GARCH (1, 1) model.