Relationship discovery

Thousands of statistical tests — without running them yourself.

Most survey projects contain far more variables than a researcher can test by hand. AddMaple's approach is different: select one column and the stats engine compares it against every other column in the project, running the appropriate test for each pair and ranking the results by statistical significance and practical effect size.

On a typical tracker or ad hoc study, that can mean thousands of pairwise tests executed in seconds — chi-square, ANOVA, t-tests, Kruskal-Wallis, Pearson correlation, and Spearman correlation — without exporting to SPSS, writing syntax, or choosing tests one at a time.

The right test is selected from the data in front of you.

You do not pick the test manually. AddMaple inspects column types and routes each comparison automatically:

• Categorical vs categorical — chi-square with Cramér's V
• Numeric vs categorical — ANOVA (3+ groups), t-test (2 groups), or Kruskal-Wallis when groups are small or non-normal, with Cohen's d or eta squared
• Numeric vs numeric — Pearson or Spearman correlation

Weighted projects use weighted versions of these tests so p-values and effect sizes respect your survey weighting scheme.

See what matters first — then inspect the evidence.

Results appear in the Stats tab, colour-coded by strength, with plain-English interpretation. Click a related column to open a pivot chart instantly, or toggle detailed calculations for p-values, effect sizes, sample sizes, and — for chi-square and ANOVA — the categories that differ most.

That is AddMaple's discovery layer: broad, automatic, and inspectable — not a black-box "insights" summary.

Shortlist before you model.

When you pick an outcome for key driver analysis, statistically associated variables surface first as candidate drivers. Relationship discovery answers what is related across the whole codebook; drivers and regression answer what matters most or what is the estimated coefficient — with a defensible starting point rather than an intuition-led search.

For significance shading in banner tables and crosstabs, see statistical testing & significance.

Keep discovery broad without making the conclusion broad.

Relationship discovery is a way to find promising directions, not to claim causation. The analyst decides which relationships are substantively plausible, checks the underlying cuts and bases, and chooses whether the next step is a driver model, regression, or focused table.