Can it help with complex multivariable models?

This "Lite" version is best for simple, univariate or bivariate analyses. For multivariable regression, survival analysis, or mixed models, use the "Statistical Method Advisor (Full)".

What if my data isn't normally distributed?

The AI will usually recommend both parametric and non-parametric options (e.g., T-test and Mann-Whitney U) and remind you to test for normality first.

Does it write R/Python/SPSS code?

By default it returns the general test name and simple command hints. Ask as a follow-up (e.g., 'give me the full R / Python / SPSS code for this analysis') and it will produce a runnable script tailored to the recommended method — specify your software in the SOFTWARE field or in the follow-up.

Should I share actual patient data or research data when describing my dataset?

No — never paste raw patient records, identifiable information, or actual data rows. Describe your dataset structurally: variable names (use generic labels if needed, e.g., "exposure_var" instead of real field names), data types, sample size, and missing patterns. This gives the AI all the context it needs without exposing sensitive data. If your institution or IRB has AI use policies for patient-linked research data, consult them first.

What is the difference between the Lite and Full versions? When should I use the Full version?

The Lite version focuses on simple comparisons with one outcome variable and two to three groups, giving concise results quickly. The Full version handles more complex scenarios: multivariable regression (adjusting for multiple confounders), repeated measures data, survival analysis, sample size calculation, and multiple primary outcomes requiring correction for multiple comparisons. If your Methods section will include Cox proportional hazards, mixed-effects models, or LASSO, use the Full version.

I have multiple outcome variables. Can I ask about all of them at once?

You can describe multiple outcomes, but be aware: testing multiple outcomes simultaneously creates a multiple comparisons problem — the more tests you run, the higher the probability of a false positive. The Lite version will flag this and suggest Bonferroni or FDR correction. However, if you have three or more primary outcomes, the Full version is better suited — it has built-in multiple comparisons guard rules that automatically calculate the corrected significance threshold.

Statistical Method Advisor (Lite) -- medPrompt

Usage Guide

Fill in your data description and research question (required); optionally specify study design, analysis software, and research field for more precise recommendations.
Click AI Run — receive a quick statistical method recommendation with rationale directly in the chat.
Ask for deeper explanation, alternative approaches, or a Methods section draft.

Medical Research Assistant

Fill in variables and run directly with AI

Wiki

What this tool does

This is a rapid statistical decision-support tool for clinical researchers. Describe your variables, group structure, and research question — the AI recommends the appropriate statistical test, checks its assumptions, and provides a ready-to-use Methods section draft. Designed for simple analyses with one primary outcome.

Decision logic

Statistical method selection depends on three questions: What type is the outcome? How many groups? Paired or independent? Common combinations:

Outcome	2 independent groups	2 paired groups	3+ groups
Continuous (normal)	Independent t-test	Paired t-test	One-way ANOVA
Continuous (skewed)	Mann-Whitney U	Wilcoxon signed-rank	Kruskal-Wallis
Binary / categorical	Chi-square / Fisher's exact	McNemar	Chi-square
Count / rate	Poisson regression	—	—

Knowing this logic helps you verify the AI's recommendation rather than accepting it blindly. The AI preserves your original data description across follow-ups — you can iterate (e.g., "what if the sample is 3 groups instead of 2") and it will explicitly note what changed and revise the recommendation accordingly.

When to upgrade to the Full version

This Lite version handles simple single-outcome comparisons. Use Statistical Method Advisor (Full) for: multivariable regression (adjusting for confounders); repeated measures or longitudinal data; survival analysis (time-to-event outcomes); sample size calculation; multiple primary outcomes requiring correction for multiple comparisons.

Data privacy

Describe your dataset structurally — variable names, types, sample size, missing data patterns. Never paste raw patient records or identifiable data.

FAQ