Vector Autoregressive Models for Multivariate Time Series Analysis on Covid-19 Pandemic in Nigeria

Abstract:

In this paper, we have been to use vector autoregressive (VAR) models for modeling and forecasting covid-19 variables with special focus on Nigeria cases from 1st march to 10th June 2020. At lag of order 2, the hypothesis of non-stationary is rejected at 5% level for all the multivariate variables using the augmented Dickey Fuller and Phillips-Perron unit root tests. The Granger causality test results indicate that there is a bivariate causal relationship among the variables by rejecting the null hypothesis of no Granger causality. The determinants of confirmed cases, new cases, and total deaths from covid-19 are generally significant at 5% level with p-value 0.0001 in each of the three derived models. The criteria AIC and log-likelihood implemented on the models confirmed that the VAR model of order 2 gives a better model for predictions and forecasts of covid-19 cases in Nigeria. This paper recommends a suitable model for handling multivariate time series data and suggests a reliable approach for forecasting future cases of covid-19 variables in the country and help health policy makers in finding solution to the unceasing upward trend in the cases of the pandemic.