Abstract
Undeniably, the Novel Coronavirus 2019, (COVID-19), has disrupted the routine functioning of the global economic and social activities. In particular, vulnerable economies such as the Small Island Developing states (SIDs) are facing unprecedented health and financial crisis. In such critical situation, some in-depth statistical models can be helpful for proper planning in terms of identifying factors that can influence significantly the number of infected COVID-19 cases and for forecasting. Modelling the COVID-19 infected series is a statistical challenge since the series are severely over-dispersed with lots of oscillations. This paper attempts a new integer-valued time series model based on the auto-regressive structure (INAR), with an oscillating Weighted Cosine Geometric (WCG) innovation term. The parameters in the proposed model constitute of the regression effects and serial auto-correlation coefficients and are estimated via likelihood and moment-based approaches. Monte Carlo simulation experiments are implemented to assess the performance and consistency of the different resulting estimators. Thereon, the INAR-WCG is applied to the COVID-19 series of various SIDs countries that include Singapore, Cape Verde, Bahrain, Mauritius and Maldives. The findings reveal that factors such as the transmission mode and the Government Stringency Index are the most influential. In terms of model fitting and forecasting, the INAR-WCG provides slightly better RMSEs than the other competing INAR-based processes.