Modernising Receiver Operating Characteristic (ROC) Curves

The justification for making a measurement can be sought in asking what decisions are based on measurement, such as assessing the compliance of quality characteristic an entity relation to specification limit, SL. relative performance testing devices and classification algorithms used is often evaluated using venerable ever popular receiver operating (ROC). However, ROC tool has potentially all limitations classic test theory (CTT) non-linearity, effects ordinality confounding task difficulty instrument ability. These limitations, inherent unacknowledged when tool, tackled here first time with modernised approach combining system analysis (MSA) item response (IRT), data from pregnancy example. new method device ability separate Rasch IRT regressions each axis curves found perform significantly better, correlation coefficients traditional area-under-curve metrics at least 0.92 which exceeds that linearised plots, Linacre’s, recommended replace other approaches assessment. resulting improved curve achieved this original should enable more reliable decision-making conformity assessment many scenarios, including machine learning, where its use metric become almost indispensable.