Title: Bayesian nonparametric inference for the three-class Youden index and its associated optimal cut-points
Authors: Vanda Inacio - University of Edinburgh (United Kingdom) [presenting]
Adam Branscum - Oregon State University (United States)
Abstract: The three-class Youden index is a generalisation of the Youden index to the case where there exist three ordinal disease classes and it serves both as a measure of diagnostic accuracy and as a criterion to choose the optimal pair of cutoff values for diagnosing subjects in practice. We develop a Bayesian nonparametric approach for estimating the three-class Youden index and its associated optimal cutoff values based on Dirichlet process mixtures, which are robust priors that can handle nonstandard features of the data, such as skewness and multimodality. A simulation study is performed and an application to data concerning the Trail Making Test Part A, which has been used to assess cognitive impairment in Parkinson's disease patients, is provided.