Approximate Bayesian inference for joint partially linear modeling of longitudinal measurements and spatial time-to-event data

Abstract: The integration of longitudinal measurements and survival time in statistical modeling offers a powerful framework for capturing the interplay between these two essential outcomes, particularly when they exhibit associations. However, in scenarios where spatial dependencies among entities are present due to geographic regions, traditional approaches may fall short. In response, this paper introduces a novel approximate Bayesian hierarchical model tailored for jointly analyzing longitudinal and spatial survival outcomes. The model leverages a conditional autoregressive structure to incorporate spatial effects, while simultaneously employing a joint partially linear model to capture the nonlinear influence of time on longitudinal responses. Through extensive simulation studies, the efficacy of the proposed method is rigorously evaluated. Furthermore, its practical utility is demonstrated through an application to real-world HIV/AIDS data sourced from various Brazilian states, showcasing its adaptability and relevance in epidemiological research.