MIXED GEOGRAPHICALLY WEIGHTED THREE PARAMETER LOG-LOGISTIC REGRESSION FOR MODELING LEPROSY INCIDENCE IN CENTRAL JAVA

Abstract

Leprosy incidence exhibited non-normal, right-skewed distribution and spatial heterogeneity, requiring a model accommodating both characteristics. This study aimed to develop and apply the Mixed Geographically Weighted Three-Parameter Log-Logistic Regression (MGWLL3R), representing the first integration of the Mixed Geographically Weighted Regression framework with the Three-Parameter Log-Logistic distribution, to identify factors influencing leprosy incidence using cross-sectionaldata from 35 regencies/cities in Central Java, based on 2024 data obtained from BPS of Central Java. Parameters were estimated using Maximum Likelihood Estimation with the Berndt-Hal-Hall-Hausman algorithm, and the optimal bandwidth was selected using minimum Cross Validation. The best model employed a Fixed Gaussian kernel with bandwidth 0.37 and achieved the lowest AICc of 90.70 compared with the LL3R and GWLL3R models. Proper sanitation access was the only global predictor, whereas unemployment, safe drinking water access, population growth, and mean years of schooling showed spatial variation. These findings support region-specific leprosy control through global and local effects.