Analysis of Features in a Sliding Threshold of Observation for Numeric Evaluation (STONE) Curve

We apply idealized scatterplot distributions to the sliding threshold of observation for numeric evaluation (STONE) curve, a new model assessment metric, examine relationship between STONE curve and underlying point-spread distribution. The is based on relative operating characteristic (ROC) but developed work with continuous-valued set observations, sweeping both observed modeled event identification simultaneously. This particularly useful predictions time series data as case much terrestrial weather space weather. identical sweep observational thresholds results in changes states quadrant boundaries shift. data-model pair's status result nonmonotonic features appear when compared an ROC same sets. Such reveal characteristics values. Many sets were created known distributions, connecting certain distinct signatures. A comprehensive suite feature-signature combinations presented, including their several other metrics. It shown that if local spread more than 0.2 full domain or bias half spread. example real-time plasma sheet electron modeling used show usefulness this technique, especially combination