
Calculate Forecast Skill Scores
skill_scores.Rd
This function calculates various skill scores and statistics for evaluating the performance of binary forecasts based on a 2x2 contingency table.
Value
A named list containing the following components:
hit: Number of hits (input)
miss: Number of misses (input)
false_alarm: Number of false alarms (input)
quiescent: Number of correct rejections (input)
EF: Event forecasts (hit + false_alarm)
NEF: Non-event forecasts (miss + quiescent)
E: Total events (hit + miss)
NE: Total non-events (false_alarm + quiescent)
SS: Sample size (total observations)
POD: Probability of Detection (hit rate)
FOM: Frequency of Misses (miss rate)
FAR: False Alarm Ratio
FOH: Frequency of Hits (hit rate among forecasts)
PON: Probability of Null (correct rejection rate)
POFD: Probability of False Detection (false alarm rate)
DFR: Detection Failure Ratio
FOCN: Frequency of Correct Null forecasts
CSI: Critical Success Index
TSS: True Skill Statistic
HSS: Heidke Skill Score
Details
The function calculates skill scores commonly used in meteorology and other forecasting domains. The contingency table structure is:
Observed
Yes No
Forecast Yes hit false_alarm
No miss quiescent
Correspondence to ML terminology:
Meteorology/Forecasting | ML/Statistics |
POD (Probability of Detection) | Recall/Sensitivity/TPR |
POFD (Prob. of False Detection) | False Positive Rate |
FAR (False Alarm Ratio) | 1 - Precision |
FOH (Frequency of Hits) | Precision/PPV |
PON (Probability of Null) | Specificity/TNR |
TSS (True Skill Statistic) | Sensitivity + Specificity - 1 |
CSI (Critical Success Index) | Jaccard Index |
HSS (Heidke Skill Score) | Cohen's Kappa (similar) |