Local BES Indicators, an Application with Mixed Effect Models and Bayesian Networks

This paper investigates the application of Bayesian Networks (BNs) and Mixed-Effect Models to analyze local well-being indicators within Equitable and Sustainable Wellbeing (BES) framework. By leveraging Conditional Gaussian Bayesian Networks (CGBNs) with random effects, the analysis effectively integrates both spatial and temporal dimensions, enhancing the estimation of missing data and accurately capturing local heterogeneity. The findings underscore the advantages of using mixed-effects models with partial pooling compared to traditional BN approaches. This method proves effective in accounting for the spatial and temporal variability inherent in well-being indicators, providing a richer understanding of how these indicators evolve over time and differ across various regions. The results not only highlight variations in well-being across communities but also facilitate a nuanced analysis that can inform informed policy-making and strategic decision processes. Future research will focus on improving predictive accuracy, exploring spatial autocorrelation, and extending the methodology to dynamic Bayesian networks.