All Times EDT
Keywords: ROC, ordinal regression, homogeneity test
In medical imaging studies with ordinal ratings, the underlying receiver operating characteristic (ROC) curve is commonly estimated using an ordinal regression model. In this talk, I will discuss a homogeneity test for covariate-adjusted ROC curves based on the ordinal regression. Moreover, pairwise comparison tests are also investigated by establishing confidence intervals of difference among the area under the curves and confidence bands of difference among ROC curves. In addition, the relationship between sample sizes and the power of the proposed homogeneity test is examined to determine the minimum required number of samples.