Sumith Gunasekera
B.Sc.(Sp.) degree in Physics in 1995 from the University of Colombo, Colpetty, Colombo 03, District of Colombo
Title: Estimating the Youden index under the multivariate ROC curve in the presence of missing values of mass diseased- and healthy-biomarker data
Biography
Biography: Sumith Gunasekera
Abstract
Abstract—In the present day scenario, classification tools have gained much attention of researchers in solving real life classification problems which arise in various disciplines. Basing on the prominence and demand to handle such problems, those variety of statistical tools have emerged and were brought under the hub of Statistical Decision Theory (SDT). The common problem of interest in classification is in allocating an individual or object to one of the predefined groups (or populations) by using a threshold. These problems were addressed by using a performance tool namely, Receiver Operating Characteristic (ROC) Curve. The Youden index (J) is a frequently used summary measure of the ROC curve. It both, measures the effectiveness of a diagnostic marker and enables the selection of optimal cut-point (c ?) for the biomarkers. The proposed research demonstrates how J for the diseased and healthy subjects can be extended to multi-biomarkers in the higher-dimensional space by analytic and extensive-computational continuation of the mass missing data of multi-biomarkers from the breast cancer patients available at http://archive.ics.uci.edu/ml/datasets/ mammographic+mass, and by intensive and extensive simulations and computations of big data with the aid of generalized variable method. This computational-extensive mass-data-based procedure is novel and reduce the high number of unnecessary biopsies by helping physicians in their decision to perform a breast biopsy on a suspicious lesion seen in a mammogram or to perform a short term follow-up examination instead. This goal is accomplished by the comparison of classical and generalized variable procedures for J and c ? for the multi-biomarkers with missing data, where missing data are cleaned or tackled by imputations. These are juxtaposed using confidence intervals, pvalues, power of the test, size of the test, and coverage probability with a wide-ranging simulation study featuring a selection of various scenarios. Index Terms—Generalized variable method, High-performing computations, Multivariate normal distribution, Multivariate Youden index, Power of the test, Size of the test, Shiftedexponential distribution, Youden index