Univariate · Break down by groups — build 2026-10-31.49
Tests → Univariate descriptive analysis now has a Break down by groups panel under the frequency table. It answers questions like Stata’s
tab ss_dir if sexo == 2 & year == 2018
sum ss_dir if sexo == 1 & year == 2018
for every combination at once: pick Rows (e.g. year) and, optionally, Columns (e.g. sexo).
| Option | What each cell shows | Stata equivalent |
|---|---|---|
| % where v = … | share of the rows with the chosen category (e.g. ss_dir = Sí) | table year sexo, statistic(mean _is) |
| Count where v = … | how many (weighted total when a weight is set) | statistic(total _is) |
| Full tabulation | every category of v as row percentages, with N | table year ss_dir sexo, statistic(percent, across(ss_dir)) |
| Mean · Sum · Std. dev. · Median · Min · Max | for numeric variables (a 0/1 variable’s mean is its share) | tabstat v, by(year) / table …, statistic(mean v) |
| N (observations) · Population (weighted) | counts | statistic(count v) / statistic(frequency) |
- Every value in the range shows every year from the lowest to the highest (2000–2024) even when a year has no data (shown as —); the chart’s lines run through those years.
- Click any cell for its full
tab v if year == 2018 & sexo == 2table (Freq., Percent, Cum.) and thesummarizeline. - Weights (Data → weight): % and means are weighted, counts and sums are weighted totals on the weight’s own scale (population estimates with an expansion factor), N is always the number of observations; min, max and median are unweighted.
- The IF subgroup applies, rows with a missing value are left out and counted in the note, and the table copies to Excel; the chart downloads as PNG (150/300/600 dpi) or SVG; Stata and R code reproduce the table.
Source: src/breakdown.js, integrated by tools/breakdown.py (run through tools/text_lab.py).
Test: node --test calculator/tests/breakdown.test.mjs.