Discrepancy Measure

A Discrepancy Measure is an evaluation metric that quantifies the difference, gap, or deviation between observed values and expected standards.

• AKA: Gap Measure, Deviation Measure, Difference Metric.
• Context:
  • It can typically identify Performance Gaps through numerical differences and statistical deviations.
  • It can typically support Quality Assessment via gap analysis and deviation scoring.
  • It can typically enable Improvement Planning through gap identification and priority ranking.
  • It can often measure Distribution Differences using divergence metrics and distance functions.
  • It can often quantify Alignment Deviation between actual outputs and target specifications.
  • ...
  • It can range from being a Small Discrepancy Measure to being a Large Discrepancy Measure, depending on its discrepancy magnitude scale.
  • It can range from being a Linear Discrepancy Measure to being a Non-Linear Discrepancy Measure, depending on its discrepancy scaling function.
  • It can range from being a Point Discrepancy Measure to being a Distribution Discrepancy Measure, depending on its discrepancy comparison scope.
  • It can range from being a Symmetric Discrepancy Measure to being an Asymmetric Discrepancy Measure, depending on its discrepancy directionality.
  • ...
• Example(s):
  • Quality Discrepancy Measures for quality gap assessment.
  • Distance-to-Gold for ideal deviation measurement.
  • Model Quality Gap for performance shortfall quantification.
  • Kullback-Leibler Divergence for distribution difference.
  • Wasserstein Distance for probability distribution gap.
  • ...
• Counter-Example(s):
  • Similarity Measure, which quantifies agreement rather than difference.
  • Correlation Metric, which measures relationship rather than gap.
  • Match Score, which identifies correspondence rather than discrepancy.
• See: Similarity Measure, Distance Metric, Divergence Measure, Gap Analysis, Evaluation Metric, Statistical Distance, Quality Discrepancy Measure, Error Metric.

---