Endometriosis is notoriously difficult to diagnosis. Patients wait an average of seven years to receive a formal diagnosis [1]. In this study, Bayesian logistic regression is applied to synthetic data to develop a predictive model which can aid in early detection of endometriosis. Comparisons are made across six candidate models with informative normal and weakly informative Cauchy priors. The Markov Chain Monte Carlo and Metropolis-Hastings algorithm is applied to these general linear models to predict the odds of possessing this gynecological condition.
KEYWORDS: Endometriosis, Bayesian inference, general linear model, Markov Chain Monte Carlo, Metropolis-Hastings, normal prior, Cauchy prior