Latent Bayesian clustering for topic modelling
in: Methodological and Applied Statistics and Demography IV - SIS 2024, Short Papers, Contributed Sessions 2, 2024.
Citation: Stival, M., Schiavon, L., Bertarelli, G., Campostrini, S. (In press) Exploring disease prevalence in Italy: a web application utilizing sample-based predictions from a Bayesian spatio-temporal model with external validation, in Methodological and Applied Statistics and Demography IV - SIS 2024, Short Papers Contributed Sessions 2 (Editors: Pollice, A., Mariani, P.), ISBN e-book: 9783031644474.
Abstract: Disease prevalence data is an invaluable resource in population monitoring due to its easy representation and immediate interpretability. However, accessing such information can often be challenging. Although, sampling methods provide an efficient alternative for inferring population prevalence and collecting correlated risk factor information, challenges arise due to potential under representation, necessitating robust modeling approaches. Hierarchical Bayesian models may offer flexible solution to integrate prior knowledge and considering individual and environmental factors in disease prevalence estimation. However, interpreting complex statistical models poses challenges for non-experts. To address this, a user-friendly web interface leveraging Shiny is proposed, facilitating comparison and validation of disease prevalence estimates. This infrastructure aims to enhance accessibility and reliability, with potential for broader applicability and future improvements.