Evaluation of the success of classification based on cut-off scores: Receiver operating characteristic curve
Keywords:sensitivity, specificity, ROC analysis, area under curve (AUC)
AbstractAim of this study to draw attention to possibilities for use ROC curve analysis (receiver operating characteristic curve) for determining the classification capabilities of the tests. Concepts of sensitivity and specificity, underlying creation of ROC curves, are explained. Interpretation of formulas for calculating the positive and negative predictive values and accuracy of the tests are also given. ROC curve is a graphical representation of sensitivity and specificity for every possible threshold score (test result) in the coordinate system where the ordinate shows the values of sensitivity and the abscissa value of 1-specificity. It is explained how to determine optimal threshold score on the basis of sensitivity and specificity, and how to perform ROC analysis in several statistical packages (SPS, PSPP and R). In the end, it is pointed to the findings within clinical psychology that are based on ROC analysis and test characteristics (such as sensitivity and specificity) on which this analysis is based.
Metrics Loading ...
How to Cite
Janičić, B., & Novović, Z. (2011). Evaluation of the success of classification based on cut-off scores: Receiver operating characteristic curve. Primenjena Psihologija, 4(4), 335–351. https://doi.org/10.19090/pp.2011.4.335-351