Member Junction
    Preparing search index...
    • Compute PSI for a continuous/numeric feature by constructing empirical quantile bins from the baseline distribution (with open boundary tails (-inf, q_1) and [q_{B-1}, +inf) and deduplicated interior edges when ties exist), tracking missing/non-finite rates in a dedicated missingness bin.

      Parameters

      • featureName: string
      • baselineValues: readonly number[]
      • evaluationValues: readonly number[]
      • numBins: number = 10
      • epsilon: number = DEFAULT_PSI_EPSILON
      • warningThreshold: number = PSI_WARNING_THRESHOLD
      • criticalThreshold: number = PSI_CRITICAL_THRESHOLD
      • minSampleSize: number = DEFAULT_MIN_SAMPLE_SIZE

      Returns FeaturePSIResult