pre-analysis-plan

name: pre-analysis-plan description: Draft pre-analysis plans by dispatching the causal-strategist agent in PAP mode. Covers AEA RCT Registry, OSF, and EGAP standards. Specifies outcomes, subgroups, multiple testing, power calculations, and exclusion rules. disable-model-invocation: true argument-hint: "[research-spec file OR topic OR 'interactive' for guided interview]" allowed-tools: ["Read", "Grep", "Glob", "Write", "Task"]

Pre-Analysis Plan

Draft a pre-analysis plan by dispatching the causal-strategist agent in PAP mode.

Input: $ARGUMENTS — path to research spec file, a topic, or interactive for guided questions.

Workflow

Step 1: Load Context

• If $ARGUMENTS is a file: read it (research spec from /interview-me)
• If $ARGUMENTS is interactive: conduct guided interview (see below)
• Otherwise: treat as topic and draft with ASSUMED placeholders marked clearly
• Read .claude/rules/domain-profile.md for field conventions

Interactive Questions (if needed)

1. What is the research question?
2. What is the study design? (RCT / natural experiment / quasi-experimental)
3. What are the primary outcome variables?
4. What is the identification strategy?
5. What subgroup analyses are pre-specified?
6. What multiple testing concerns exist?

Step 2: Launch causal-strategist Agent (PAP Mode)

Delegate to the causal-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

Step 3: Platform-Specific Adaptation

• AEA RCT Registry: Most structured. All fields required. Register before intervention.
• OSF: More flexible. Good for observational studies and natural experiments.
• EGAP: Development economics focused. Additional governance questions.

For observational studies: adapt template (identification strategy replaces randomization, comparison group replaces control, add identification assumption discussion).

Step 4: Launch econometrics-critic Review (Optional)

If the PAP includes a complex identification strategy, optionally dispatch the econometrics-critic:

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.

Step 5: Present to User

CRITICAL: Flag every ASSUMED item clearly. The researcher must review and approve before registration. A registered PAP with errors is worse than no PAP.

Principles

• Pre-specification is the point. Everything decided before seeing outcomes.
• Be honest about what's exploratory. Label subgroups and secondary outcomes clearly.
• Power calculations require assumptions. State every assumption. Show sensitivity.
• This is a commitment device. Make sure the researcher understands what they're committing to.
• causal-strategist creates, econometrics-critic critiques (optional but recommended for complex designs).