Decision quality under uncertainty

Purpose

ws4.opt.decision_quality evaluates a set of decisions across uncertainty scenarios. It reports expected outcome, worst case, regret against the per-scenario best, rank stability, and optional constraint satisfaction.

Contract

Input is a long table with:

  • a decision column;

  • a scenario column;

  • a finite, lower-is-better outcome column;

  • an optional boolean/0-1 feasibility column.

Decision/scenario pairs must be unique. Scenario weights default to equal weights.

Outputs

  • expected outcome;

  • worst case and best case;

  • expected regret and maximum regret;

  • rank range and rank stability;

  • weighted constraint satisfaction.

The deterministic, robust, and minimax-regret choices are reported separately. Ties are allowed.

Containment example

The P203 closure matrix is used as a worked example: five closure times across five weather members. The earliest closure times dominate; later closure has higher expected area, higher worst case, and lower rank stability.

This is a synthetic containment diagnostic, not an operational suppression decision. See notes/reports/decision-quality-containment.md.