Group Difference Analysis

Automatically test inter-group differences and surface the variables that matter.

From normality checks to post-hoc and multiple-comparison correction — all in one flow. Results export directly into publication-ready paper format.

At a glance

Common use

Clinical trials · A/B tests · Education · Policy

Outcome variable

1 categorical column (e.g. treatment / control)

Input data

CSV / XLSX (≤ 30MB), rows = samples, columns = variables

Sample size

At least 3–5 per group recommended

Plan

Free on the FREE plan

Data preparation

  1. 1Tabular file (CSV / XLSX, up to 30MB)
  2. 2Rows = individual observations, columns = variables
  3. 3Numeric (age, blood pressure) and categorical (gender, group) can coexist
  4. 4Outcome: 1 categorical column with ≥ 2 values
  5. 5≥ 3–5 samples per group recommended for stable inference

If your outcome is numeric (e.g. 0/1, scores), convert it to Categorical in the 'Modify or remove variables' step so group tests run correctly.

Workflow

  1. 1Variable cleanup (auto-exclude + user review)
  2. 2Missing-value imputation (mean / median / mode / user-defined)
  3. 3Encoding (One-Hot / Ordinal)
  4. 4Pick the outcome variable
  5. 5Statistical tests (normality → parametric / non-parametric auto)
  6. 6Post-hoc + multiple-comparison correction
  7. 7Visualisation (boxplot / histogram / correlation)
  8. 8Regression modeling (linear / logistic / Firth)

Supported analyses

  • Descriptive statistics

    Compare group means · SDs · medians at a glance

  • t-test / Mann-Whitney U

    Two-group difference — auto-picked by normality

  • ANOVA / Kruskal-Wallis

    Three+ group difference + post-hoc

  • Chi-square test

    Independence between two categorical variables

  • Linear / Logistic / Firth regression

    Per-variable effect (stepwise supported)

  • ROC + AUC

    Discriminative power of classification models

Use cases

  • Clinical trial: treatment effect

    Compare blood pressure / cholesterol between treatment vs control via t-test / Mann-Whitney.

  • A/B test: design comparison

    Test click-rate / time-on-page differences via t-test; conversion-rate via chi-square.

  • Education effect

    Compare pre/post scores per group with multiple-comparison correction for credibility.

What you get

  • Test result tables (statistic, p-value, effect size)
  • Group boxplots / histograms
  • Post-hoc matrices + visualisations
  • Regression coefficients + CIs + diagnostic plots
  • Auto-generated paper (LaTeX → PDF)