Generalized Linear Mixed-Model Tree for Modeling Dengue Fever Cases

The GLMM tree demonstrates flexibility when applied to complex dataset structures such as multilevel and longitudinal data. However, there has been no assessment of the performance of GLMM trees on panel data structures. This study aims to assess the performance of the GLMM tree on a panel data structure using a case study of dengue fever cases in West Java. The performance evaluation focuses on the accuracy of the model. The dataset includes cross-sectional data from 27 regencies/cities in West Jawa, covering different regions at a single point in time, and time-series data from 2014 to 2022, tracking dengue fever cases over the years. The results of this study show that the GLMM tree model is suitable for panel data that exhibit nuanced or intricate variability unrelated to temporal effects. When developing the incidence rate of the dengue fever model, the GLMM tree separates into two submodels depending on a GRDP growth rate threshold of 5.5%. The GLMM tree model shows significant differences in the incidence rate of dengue fever between regencies/cities. However, the differences in the incidence rate of dengue fever from year to year between the regencies/cities are not significant. It indicates that local factors, such as research predictor variables, are more dominant in influencing the incidence rate than global factors.