Draft pre-analysis plans by dispatching the Strategist agent in PAP mode. Covers AEA RCT Registry, OSF, and EGAP standards. Specifies outcomes, subgroups, multiple testing, power calculations, and exclusion rules.
Draft a pre-analysis plan by dispatching the Strategist agent in PAP mode.
Input: $ARGUMENTS — path to research spec file, a topic, or interactive for guided questions.
$ARGUMENTS is a file: read it (research spec from /interview-me)$ARGUMENTS is interactive: conduct guided interview (see below).claude/rules/domain-profile.md for field conventionsDelegate to the strategist agent via Task tool:
Prompt: Draft a pre-analysis plan for "[topic/spec]".
Mode: PAP (pre-analysis plan).
Platform: [AEA RCT Registry / OSF / EGAP — ask user if unclear].
Include all standard sections:
1. Study overview (question, design, treatment, control)
2. Outcomes (primary, secondary, mechanism — with measurement details)
3. Estimating equations (with notation protocol)
4. Subgroup analyses (pre-specified, with justification)
5. Multiple testing correction (Bonferroni / BH / Romano-Wolf)
6. Power calculations (MDE, baseline stats, sample size)
7. Sample and exclusion rules (inclusion, attrition, outliers)
8. Data and analysis (sources, software, randomization)
9. Timeline
10. Deviations log (empty template)
Save to quality_reports/pre_analysis_plan_[topic]_[date].md
For observational studies: adapt template (identification strategy replaces randomization, comparison group replaces control, add identification assumption discussion).
If the PAP includes a complex identification strategy, optionally dispatch the Econometrician:
Prompt: Review the pre-analysis plan at [path].
Focus on: identification assumptions, estimator choice, power calculation assumptions.
Flag any specifications that might be problematic before registration.
CRITICAL: Flag every ASSUMED item clearly. The researcher must review and approve before registration. A registered PAP with errors is worse than no PAP.