Show significances (group comparisons) in tables and bar/column charts

You can display significances in tables, column charts and bar charts.

For a chart, select the test you want under Settings → Significances.

The following tests are currently available

• Chi-square (complement comparisons):
  Compares expected vs. observed frequencies between each individual cell and its complement (all other categories).

• Chi-square (pairwise comparisons):
  Compares expected vs. observed frequencies between cells in a pairwise manner.

• Z-test (proportion comparisons):
  Compares the percentages between cells pairwise for significance

• t-test (independent samples; pairwise):
  Compares means pairwise between two independent groups (e.g. gender) with respect to a dependent variable.

• t-test (repeated measures):
  Compares means pairwise between two measurement points (or conditions) for the same individuals with respect to a dependent variable.

• Chi-square (independence)
  Tests two variables for statistical independence.

• In progress: t-test (dependent samples)
  Compares means pairwise between two ratings (e.g. products) for the same individuals with respect to a dependent variable.

Recommended metrics

• Chi-square tests: absolute values

• Z-test: percent

• t-test: mean

Notes:

• For the t-tests, the dependent variable must be defined as numeric in the codebook!

• For the chi-square tests, the Yates correction can be enabled/disabled in the chart settings.

Multiple-testing correction

When you compare many columns at once, some differences look significant by chance alone. From release 2026.3.200, DataLion can correct the p-values for this. Open Chart style → Customize → Significances and choose under Multiple-testing correction:

• None: the default, no correction.
• Bonferroni: the strictest correction.
• Holm: guards against chance findings as well as Bonferroni but is less strict, so it flags more differences.
• Benjamini-Hochberg (false discovery rate): limits the share of false findings among the significant results.

DataLion corrects each table row for the number of comparisons in that row. The legend names the method, for example “Pairwise, Holm-adjusted per row.”, on screen and in every export. The chi-square independence test has no correction.

Significances on weighted data

From release 2026.3.200, z-tests and chi-square tests on weighted shares use Kish’s effective sample size as their base. It gets smaller the more uneven the weights are, so the tests flag fewer differences that only look significant because of the weighting. The legend then says “Weighted: tested on the effective sample size (Kish).”, and charts and tables show the figure as n eff. next to the base. More in Weighting in DataLion.

How significances are displayed (significance levels can be found in the legend):

• Complement comparisons: asterisks indicate whether one group differs from the others. The number of asterisks represents the significance level.

• A letter after a value indicates which comparison group the value is significant against. Each comparison group is assigned a letter (either in the columns or in the labels for the categories). The way the letter is displayed represents the significance level.