Comparison of Poisson and Negative Binomial Regression Models in Identifying Factors Influencing Covid-19 Deaths in Indonesia.

This research compares Poisson Regression and Generalized Negative Binomial (GNB) Regression to underscore the factors that influence the growth of COVID-19 deaths in Indonesia. Count data such as mortality cases often violates the Poisson assumption of equidispersion (null mean equals variance) causing overdispersion. The GNB model is suggested as a remedy for overdispersed data crime prevention has become increasingly necessary for systematic development because secondary data from the Indonesian government has included dependable variables such as mortality rates for people aged over 60, diabetes mellitus, heart disease, lung disease, healthcare worker percentages, referral hospitals, and the population. The Poisson Regression reported R² of 87.67% and experienced overdispersion (θ₁ = 356.27, θ₂ = 417,597). The GNB model, in contrast, with a lower AIC (499.5566), overtook Poisson. Important factors that had significant impact on both models were mortality rates for individuals over 60, diabetes mellitus, healthcare workers, and referral hospitals, whereas heart and lung disease mortality rates were the ones that were not material. The GNB model had a better fit and tackled the issues of overdispersion in the Poisson Regression.