Testing an Underlying Assumption on a ROC Curve Based on Rating Data

Publisher Website
Google Scholar
Abstract
A fully Bayesian approach to a general nonlinear ordinal regression model for ROC- curve analysis is presented. Samples from the marginal posterior distributions of the model parameters are obtained by a Markov-chain Monte Carlo (MCMC) technique— Gibbs sampling. These samples facilitate the calculation of point estimates and cred ible regions as well as inferences for the associated areas under the ROC curves. The analysis of an example using freely available software shows that the use of nonin formative vague prior distributions for all model parameters yields posterior summary statistics very similar to the conventional maximum-likelihood estimates. Clinically im portant advantages of this Bayesian approach are: the possible inclusion of prior knowl edge and beliefs into the ROC analysis (via the prior distributions), the possible cal culation of the posterior predictive distribution of a future patient outcome, and the potential to address questions such as: "What is the probability that a certain diagnostic test is better in one setting than in another?" Key words: ROC curve; diagnostic test; ordinal regression; Bayesian methods; MCMC; Gibbs sampling; maximum likelihood (Med Decis Making 1998;18:436-443)