Statistical Methods

Audit Feasible One-to-One Assignments Before Allocating Slots

Define the task

Use this check when anonymized tasks must be paired with eligible slots and neither side can be reused. Construct a rectangular matrix with items as rows and slots as columns. Mark 1 only where a documented eligibility rule allows the pair. Save row and column labels separately. This is an operational feasibility calculation; it is not propensity-score matching, random allocation or a fairness assessment.

Carry out the check

First validate eligibility, then maximize the number of pairs. Inspect unassigned items and constraints that prevent full allocation. Multiple maximum matchings can exist, so do not interpret the returned pairing as uniquely preferred. If costs or priorities matter, specify a separate optimization objective. The browser tool treats zero as no edge. In SciPy sparse input, explicitly stored zero entries still represent edges: eliminate unintended stored zeros before comparing implementations.

Worked hypothetical example

Hypothetical example: three tasks have eligibility rows [1,1], [1,0], [0,1] for two slots. At most two pairs are possible because only two slots exist; assigning task 1 to slot 0 and task 2 to slot 1 achieves that bound. Task 0 remains unmatched. A different two-pair result can be equally valid. Report the maximum count, unmatched labels and tie policy, rather than only the selected pairs.

Method checked 22 November 2026 against Primary method documentation: maximum_bipartite_matching.

Open the eligibility maximum matching.

Download the eligibility assignment audit worksheet.

Related reading

Reading path: Analyze data. Guide 193 of 202. Choose any guide to suit your project.

Statistical Methods

Review Active Bounds in a Constrained Linear Fit

Practical steps and a hypothetical example: Audit coefficient bounds and residuals in a constrained linear fit.

Statistical Methods

Check a Graph Smoothing Penalty Against Its Edge Definitions

Practical steps and a hypothetical example: Measure disagreement across a weighted undirected graph.

Browse this reading pathTools for this stageGuides and downloads for this stage