PEMODELAN GEOGRAPHICALLY WEIGHTED RIDGE REGRESSION PADA KASUS TUBERKULOSIS DI INDONESIA

Abstract

Tuberculosis remains a major public health issue in Indonesia, with substantial variation in case notification rates (CNR) across provinces, indicating the presence of spatial heterogeneity and potential multicollinearity among influencing factors. This study aims to model CNR of tuberculosis in Indonesia in 2024 using the Geographically Weighted Ridge Regression (GWRR) method, which simultaneously addresses spatial heterogeneity and multicollinearity. Secondary data from 38 provinces were analyzed, including variables such as population percentage, GERMAS implementation, smoking prevalence, HIV cases, poverty rate, hospitals, and malnutrition among infants. The analysis began with Ordinary Least Squaress (OLS), followed by diagnostic tests revealing heteroskedasticity and multicollinearity, justifying the use of GWRR. The results show that GWRR produces stable local parameter estimates across provinces with a high coefficient of determination (97%) and relatively low RMSE, indicating strong model performance. Spatial analysis reveals that dominant factors vary by region, with malnutrition and smoking showing strong influence in several provinces. Overall, GWRR proves effective in capturing spatial variation and improving model stability in tuberculosis analysis.