pmi-coverage-deviation - Audit category–sub-topic coverage deviations
Computes PMI and related residuals from a dense 2D count table to identify over- and under-represented category–sub-topic pairs.
Tags
Updated: 2026-09-28Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Compute marginal proportions
- Calculate independence baselines
- Compute smoothed PMI
- Normalize PMI scores
- Rank coverage deviations
- Audit differential coverage
Inputs
- 2D non-negative count matrix
- Category labels
- Sub-topic labels
- Smoothing parameter
- Minimum count threshold
Outputs
- PMI matrix
- NPMI matrix
- Ranked over-coverage pairs
- Ranked under-coverage pairs
- Differential coverage map
- Optional compact state