Modelling correlated proportions
Statistical models, parameter inference, and diagnostics for pairs of correlated proportions.
With Luiz Max Carvalho, I study statistical models for pairs of correlated proportions. Starting from a bivariate beta distribution, this project develops classical and Bayesian estimation methods, prior elicitation, and diagnostics for model adequacy. It also compares beta and logistic-normal models using simulations and data on childhood vaccination coverage and diagnostic-test sensitivity and specificity.
The resulting paper, Modelling correlated proportion data: Bayesian and classical approaches to inference and diagnostics, was published in TEST (2026); the link provides the accepted manuscript. Bivariate beta distribution: parameter inference and diagnostics is an earlier version.