Comparison of Poisson Inverse Gaussian (PIG) and Negative Binomial (NBR) Regression Models Using Generalized Linear Models (GLM) in Dengue Fever Cases in East Java Province in 2024

Nindy Aulia Eka Puspita

Ambidextrous : Journal of Innovation, Efficiency and Technology in Organization
Vol. 4 No. 01: Ambidextrous: Journal of Innovation, Efficiency and Technology in Organization

Background: Dengue Hemorrhagic Fever (DHF) is a major health problem in East Java Province due to population density, sanitation, availability of medical personnel, and altitude. DHF count data often experiences overdispersion, challenging the Poisson regression assumption in the Generalized Linear Model (GLM). Objective: To compare the performance of Poisson, Negative Binomial (NBR), and Poisson Inverse Gaussian (PIG) ​​models in modeling DHF cases. Method: A cross-sectional quantitative study analyzed 38 districts/cities in East Java (saturation sampling) using secondary data processed with . Analysis: Descriptive statistics, multicollinearity test (VIF<10), overdispersion test, and model comparison via AIC/deviance. Results: NBR achieved the lowest AIC (412.10) compared to Poisson (1466.60) and PIG (412.41), confirming significant overdispersion (deviance/df ≈37.16). Population density and altitude had a significant effect (p<0.05). Conclusion: NBR is optimal for DHF modeling, supporting targeted environmental and health interventions in East Java.