Analyzing Covid-19 Infection Patterns in Nigeria's Geopolitical Zones: A Comparative Study Using Count Distribution Models

Abstract

Count data regression is a commonly employed technique in various fields, particularly within the health sector, where precise model selection depends on the characteristics of the response variable. While continuous response variables are typical in regression analyses, significant attention has been devoted to modeling discrete variables. Notably, existing literature has underscored the potential spatial variance in COVID-19 incidence counts within Nigeria. Numerous scholars have explored alternative estimation methods for COVID-19 cases, motivating this study to investigate frequency disparities among different regions in Nigeria. The primary aim of this study is to compare the prevalence of COVID-19 infection cases and identify the optimal model for characterizing infection case prevalence across geographical zones in Nigeria. To achieve this, five statistical models were assessed: Poisson, Negative Binomial, Generalized Poisson, Zero-Inflated Poisson, and Zero-Inflated Negative Binomial (ZINB) distributions. Model selection criteria, including AIC, P-values, and Chi-square values, were employed for each infection case in every zone. Both ZINB and Negative Binomial models consistently address over-dispersion. However, ZINB, by accounting for excessive zeros, emerges as the optimal choice, providing accurate insights into pandemic dynamics across diverse regions in Nigeria. These findings offer actionable guidance for combating COVID-19, with ZINB models consistently demonstrating effectiveness in representing infection data across Nigeria's various regions, effectively managing over-dispersion and excess zeros.