Regression and Medical Test Accuracy
It is well known that test accuracy depends on many factors, including diﬀerences in readers and diﬀerences in various patient characteristics. For example, the age of the patient, their gender, the stage of the disease, and the therapy received, all have a bearing on the measured test accuracy. This chapter describes Bayesian regression procedures for estimating the eﬀect of patient and reader covariates on test accuracy, as measured by classiﬁcation probabilities, predictive probabilities, and diagnostic likelihood ratios. In the case of quantitative diagnostic scores, regression techniques will be used to allow for these patient and reader characteristics when estimating the receiver operating characteristic (ROC) area. For additional information on this, refer to Chapters 3 and 6 of Pepe  and Chapter 8 of Zhou et al. .