Modelling the Number of Stunting Cases in Indonesia in 2022 Using Negative Binomial Regression to Address Overdispersion

Abstract

This study models the incidence of stunting in toddlers in Indonesia in 2022 using negative binomial regression to address the overdispersion issue often present in count data. The Poisson regression model, typically used for count data, showed less accurate results due to the variance exceeding the mean, indicating overdispersion. By adopting a negative binomial regression approach, this study accommodates higher variability in the data, leading to more accurate estimates. The results reveal that the percentage of pneumonia cases and low birth weight are significant factors in stunting incidence. In contrast, other variables, such as complete basic immunization and poverty levels, are insignificant in the final model. The final negative binomial model yielded a lower AIC value than the initial model, indicating an improved model fit, with an R-squared (Nagelkerke's R²) of 50.50%. This study offers enhanced insights into the factors influencing stunting, supporting more targeted health policy decisions to reduce stunting rates in Indonesia.