---
name: specify-metrics
description: "Define primary/secondary metrics, estimands, directionality, uncertainty reporting, and decision thresholds before analysis."
---

# specify-metrics

## Purpose

Define primary/secondary metrics, estimands, directionality, uncertainty reporting, and decision thresholds before analysis.

## Input contract

```yaml
required: [estimand, outcome_schema, decision_thresholds]
optional: [evidence, assumptions, prior_results]
constraints: [use named scientific objects; retain provenance and missingness; $\alpha$ = 0.05 and power = 0.8 where applicable]
```

## Procedure

1. Validate the typed inputs and state the decision this operation must support.
2. Apply the declared operation to the named object; record intermediate values that affect interpretation.
3. Check boundary conditions and counterexamples, then emit the result with uncertainty and source links.

If estimands, metrics, and decision thresholds are fixed, consider `estimate-sample-size` as the next tactic.

## Output contract

```yaml
produces: [specify_metrics_result, evidence_trace, uncertainties]
delta_fields: [evidence_updates, uncertainties]
```

## Quality gates

- Inputs are named scientific objects with compatible schemas.
- Every material result has a derivation or source reference.
- Fixed statistical criteria remain exact where applicable: $\alpha$ 0.05 and power 0.8.

## Failure and counterexamples

Return a failed operation with the violated precondition when inputs are incomplete, assumptions are unsupported, or a counterexample defeats the result.

## Provenance map

- intermediate: experiment-execution/metric-specification