Affiliation:
1. Department of Statistics Bahauddin Zakariya University, Multan, PAKISTAN
Abstract
The receiver operating characteristic (ROC) curve is a well-known graphical method to describe the accuracy of a diagnostic test. In this paper, Logistic kernel is proposed with its optimal bandwidth and mean squared error. To observe the performance of our proposed kernel estimator, the comparison is made with a Gaussian kernel by using different bandwidths and ROC curve and the area under the curve (AUC) are calculated. For illustration, Kidney cancer data is used and the logistic kernel is found more pragmatic and sensitive biomarker to detect Kidney cancer. The outstanding performance of logistic kernel is also observed in simulation studies and we recommend using nonparametric ROC curve using logistic kernel.
Publisher
World Scientific and Engineering Academy and Society (WSEAS)
Reference35 articles.
1. Lusted LB, Logical Analysis in Roentgen Diagnosis: Memorial Fund Lecture, Radiology, Vol.74, No.2, 1960, pp. 178– 193.
2. Zangiacomi E, Louzada-Neto F, Braganca B, A Curva ROC Para Testes Diagn´osticos, Cadernos Sa´ude Coletiva, Rio de Janeiro, Vol.11, No.1, 2003, pp. 7–31.
3. Zhou XH, McClish DK, Obuchowski NA, Statistical Methods in Diagnostic Medicine, John Wiley & Sons, 2011.
4. Zhou XH, Harezlak J, Comparison of Bandwidth Selection Methods for Kernel Smoothing of ROC Curves, Statistics in Medicine, Vol.21, No.14, 2002, pp. 2045–2055.
5. Swets JA, Form of Empirical ROCs in Discrimination and Diagnostic Tasks: Implications for Theory and Measurement of Performance, Psychological Bulletin, Vol.99, No.2, 1986, pp. 181.