TY  - JOUR
TI  - Evaluating the Performances of ROC Curve EstimationMethods for Different Distributions andDifferent Kernel Functions
AB  - Objective: Receiver operating characteristic (ROC)curve is a statistical method used to examine the actual effective ness of a diagnostic test or a biomarker in a comprehensive andreliable way. Several methods have been proposed to estimate ROCcurve properly. The aim of the present study is to compare recent ROC curve estimation methods for different distribution and sam ple sizes. Material and Methods: Log-concave density andsmooth log-concave density estimate based ROC curve estimation,kernel based ROC curve estimation with Gaussian, Epanechnikov,rectangular, triangular kernels, and binormal ROC estimationmethods were compared for different simulation scenarios. Re sults: The ROC curve estimation methods based on kernel esti mates gave their best performances when the biomarker values ofnon-diseased group are normal but the biomarker values of the dis eased group are right-skewed, with a notable difference from othermethods. Epanechnikov and rectangular kernel methods yieldedbetter performance than other kernel methods in small sample sizes;but this difference disappeared as the sample size increased. Themethods based on kernel or log-concave density estimate gave theirworst results for the simulation scenario where the data were non normal but symmetric. Conclusion: The performances of the othermethods examined in the study exceeded the performance of thebinormal method in highly skewed data in both groups and whenthe distribution of diseased and non-diseased populations wereright-skewed and normal, respectively.
AU  - Sigirli, Deniz
DO  - 10.5336/biostatic.2021-85803
PY  - 2021
JO  - Türkiye Klinikleri Biyoistatistik Dergisi
VL  - 13
IS  - 3
SN  - 1308-7894
SP  - 225
EP  - 235
DB  - TRDizin
UR  - http://search/yayin/detay/503911
ER  -