How to Ask AI Methodology Questions
Ask useful questions about ATE, AMCE, clustered standard errors, interactions, restrictions, balance, and experimental design.
The Short Version
Ask useful questions about ATE, AMCE, clustered standard errors, interactions, restrictions, balance, and experimental design. A practical way to begin is to name the design: conjoint, survey experiment, A/B test, or ordinary questionnaire, then state the randomized unit, treatment or attributes, and primary outcome. Finish by making sure you confirm publication-critical choices with the cited documentation, exported scripts, and appropriate methodological review. The full walkthrough below explains each part in order.
Is This Guide for You?
Researchers who understand their substantive question but need a clear explanation of the method or estimand.
Follow These Steps
Work through these 6 steps at your own pace. The wording is intentionally practical, and you can return to the checklist at the end when you are ready to review your work.
- Name the design: conjoint, survey experiment, A/B test, or ordinary questionnaire.
- State the randomized unit, treatment or attributes, and primary outcome.
- Ask about one concept or decision at a time.
- Include restrictions, repeated tasks, subgroup plans, or missingness when they affect the estimand.
- Ask what the estimate does and does not identify.
- Confirm publication-critical choices with the cited documentation, exported scripts, and appropriate methodological review.
What This Helps You Accomplish
A technically correct answer depends on the design. Giving the assistant a compact description of assignment, outcomes, and estimands produces more useful guidance than asking a broad statistics question without context.
A Quick Confidence Check
- Name the design.
- Identify treatment and outcome.
- State the reference category.
- Mention restrictions and clustering.
- Separate confirmatory and exploratory analysis.
- Verify high-stakes decisions.
Common Questions
Can I use this guide if I am new to this?
Yes. Researchers who understand their substantive question but need a clear explanation of the method or estimand. Follow the steps in order, start with a small test, and use the final checklist before you field or report anything important.
What is the simplest way to get started?
Begin by name the design: conjoint, survey experiment, A/B test, or ordinary questionnaire. Next, state the randomized unit, treatment or attributes, and primary outcome. You do not need to perfect every setting before running a small preview or pilot.
How do I know when I am ready?
Use the confidence check above. In particular: Name the design. Identify treatment and outcome. State the reference category. Mention restrictions and clustering. When the decision is consequential, keep your study documentation and ask a qualified colleague or methods reviewer to examine the design as well.
Related Guides
Research artifact: sample data, codebook, and scriptsResearch Pathways
Product names and fielding platforms mentioned in this guide belong to their respective owners. Conjoint Survey is an independent academic research tool and is not affiliated with, sponsored by, or endorsed by those services.