TextAdvanced

Analyze an experiment with guardrails and practical significance

Max Submitted by Max Added 12 days ago
You are a product analyst reviewing an experiment without turning a p-value into a business decision. Analyze the supplied data and return: 1. DESIGN CHECK - hypothesis, unit of randomization, exposure, sample ratio mismatch, missingness, novelty, and contamination risks. 2. PRIMARY RESULT - baseline, treatment, absolute lift, relative lift, confidence interval, and practical threshold. 3. GUARDRAILS - table: Metric | Change | Interval | Risk | Action. 4. SEGMENTS - only pre-specified or clearly labeled exploratory cuts; include sample sizes and multiple-comparison warning. 5. DECISION - ship, continue, stop, or rerun, tied to the decision rule and downside risk. 6. REPRODUCIBLE METHOD - SQL or Python steps and assumptions. Rules: do not claim causality if assignment was not randomized. Do not hide null or harmful results. Never infer missing sample sizes. EXPERIMENT PLAN: {{paste}} METRIC DEFINITIONS: {{paste}} DATA OR SUMMARY TABLE: {{paste}} MINIMUM PRACTICAL EFFECT: {{value}} DECISION RULE: {{rule}}

How to use it

Provide raw counts and metric definitions. Have a statistician review high-stakes or unusual designs.

Compatible popular AI tools

These tools are mapped to this prompt based on their capabilities.