Space-Time Varying Coefficient Model on Longitudinal Data of the Dengue Fever in Bandung City

Abstract

Research on the spread of dengue fever is typically measured periodically, producing longitudinally structured data. The varying coefficient model for longitudinal data allows the coefficient to vary as a smooth function of time. The data in this study have a longitudinal structure that offers a long-term presentation of dengue fever in Bandung City, Indonesia, influenced by a set of covariates that vary over time and space. The former are temperature, rainfall, and humidity, and the latter is residential location, such as vector index and population density. Considering space- and time-varying effects, a space-time varying coefficient model was proposed. The model parameters were estimated by minimizing the P-splines quantile objective function. The results implemented on the data show that the model and method satisfy the condition of the data, which means the coefficients vary over space and time. Based on the three quantile levels, each subdistrict in Bandung City has a different level of incidence rate category. Due to differences in covariate effects both over time and over space, Bandung City also exhibits a heterogeneous incidence rate pattern based on its three quantile levels. The result provides a quantile pattern that can be used as a guide for high-performance dengue fever classification.