A derived variable is a new column calculated from an existing one. Use it to test ideas such as "does spending affect conversions a week later?" (lag) or "is the relationship straighter on a log scale?" (log). The new variable appears everywhere in the app, marked derived.
None yet.
Pick two variables. 4CorrLab chooses the right measure from their types: r and ρ for two continuous variables, η for a continuous and a categorical one, Cramer's V for two categorical ones.
See many relationships at once. Tick the variables you want (at least 2). You get up to two tables: one for the continuous variables (Pearson r below the diagonal, Spearman ρ above it, so you can compare them cell by cell), and one for every pair that involves a categorical variable (η or Cramer's V). Click any cell to study that pair in Correlate.
Compare an outcome across groups twice: once pooled (all rows together) and once split by a third variable. If the winner changes when you split, the third variable is a confounder and the pooled comparison is misleading (Simpson's Paradox). Try the Simpson's Paradox demo dataset with Outcome = converted, Group = channel, Split by = account_size.
Significance level used throughout: 95% (change it in the Correlate tab).
The business question, the data you used and what you set out to test. It opens the report.
Everything you added with "Add to Report", in order. You can add or edit your notes on each finding here at any time.
What the findings mean together and what you would recommend. It closes the report.
This is exactly what the downloaded report contains. It is always shown on a white page so it prints cleanly.