Friday, April 10, 2020 - 10:00am
Join us online as the Department of Biostatistics and Epidemiology welcomes Dr. Leonidas Bantis from the University of Kansas Medical Center's Department of Biostatistics & Data Science. Bantis will deliver a talk titled "Comparison of correlated biomarkers at affordable operating points in trichotomous settings with applications to prostate cancer" on Friday, April 10 beginning at 10:00 AM via Zoom.
Join Zoom Meeting: https://umass-amherst.zoom.us/j/231899930
Meeting ID: 231 899 930
One tap mobile
+13126266799,,231899930# US (Chicago)
+16468769923,,231899930# US (New York)
When evaluating a new biomarker clinical investigators may be interested in avoiding over- or under-diagnosis of a disease, as opposed to exploring general accuracy measures. Comparisons of two competing biomarkers under such a focused setting may be challenging, especially when dealing with three progressive stages of the underlying disease status. The receiver operating characteristic (ROC) curve is typically employed when one wants to evaluate the discriminatory capability of a continuous or ordinal biomarker in the case where two groups are to be distinguished, commonly the “healthy” and the “diseased.” In cases where three groups are available, the ROC surface can be employed. We will discuss new methodologies for comparing two continuous biomarkers that refer to a trichotomous disease status, when both markers are applied to the same patients. We will focus such comparisons at clinically affordable operating points, which are more relevant to patients and physicians. We will explore delta‐based parametric techniques, power transformations to normality, and bootstrap‐based smooth non-parametric techniques to investigate the performance of an appropriate test. We illustrate our approaches using a real data set that refers avoiding over-diagnosis of prostate cancer.