Meta-Analysis: Subgroup (Q-between)

Verified

Meta (legacy hub)

Independently verified. Every statistic this test reports has been re-derived against an independent reference — never the library the pipeline itself calls — the rendered output was read back in a browser, and the result is locked with a committed regression suite.

Run this test straight away on a free built-in teaching dataset — no data of your own needed — or bring your own. Either opens the guided workspace: variable setup, assumption diagnostics, results with effect sizes and confidence intervals, figures, and APA-ready reporting.

Loading teaching datasets…

Or use your own dataset

Loading your datasets…

Tests whether the pooled meta-analytic effect differs across study subgroups.

Splits studies into subgroups (e.g. by dose or population) and tests whether their pooled effects differ (Q-between), explaining part of the heterogeneity.

Worked example

Does the treatment effect differ between adults and children?

18 trials were pooled and split by age group; a subgroup analysis tests the between-group difference.

Result

The effect was larger in adults (SMD = 0.51) than children (0.19), Q-between(1) = 11.1, p = .001.

How you'd report it (APA)

Subgroup analysis showed a larger effect in adults than children, Q-between(1) = 11.1, p = .001.

Try it yourself: Load this ready-made sample and follow the run above.

When to use it

  • A categorical study-level moderator
    Studies split naturally into groups. e.g. adult vs paediatric trials, high vs low dose.
  • You want to explain heterogeneity
    Heterogeneity is present and you have an a-priori reason it might differ by group.
  • Enough studies per subgroup
    Each subgroup is itself a small meta-analysis and needs several studies to pool credibly.

When NOT to — use instead

  • The moderator is continuous
    Carving a continuous variable (dose, year) into bins throws away information and depends on arbitrary cut-points. → Meta-regression
  • Very few studies overall
    Splitting a small set into subgroups leaves each too sparse to pool; the Q_between test has almost no power.

Hypotheses

H₀: the pooled effects of the subgroups are equal — Q_between = 0, i.e. the moderator does not explain any of the heterogeneity.
Hₐ: at least one subgroup's pooled effect differs. Tested by Q_between on (number of subgroups − 1) df; a significant result means the grouping accounts for part of the between-study variation.

Parameter tested: the difference between subgroup pooled effects, tested via Q_between; each subgroup also has its own pooled effect and CI

Assumptions (and what to do if they fail)

Subgroups defined in advancehigh

Check: The grouping is pre-specified, not chosen after seeing which split gives p < .05.

If violated: Post-hoc subgrouping is a fishing expedition; any 'significant' difference is likely spurious.

Correct Q_between referencehigh

Check: Q_between should compare each subgroup pool against the overall weighted grand mean (Borenstein), not against an arbitrary reference subgroup.

If violated: A wrong reference distorts the between-groups statistic — this exact bug was fixed in the runner; the note documents why the reference matters.

Enough studies in each subgroupmedium

Check: A subgroup of 1–2 studies has an unstable pooled effect and τ².

If violated: A 'difference' can be an artefact of one small subgroup's noise.

Between-study, not within-study, comparisonmedium

Check: Subgroups are compared across studies, so the contrast is observational even when the studies are randomised trials.

If violated: A subgroup difference is suggestive, not causal — it is confounded by everything else that differs between the groups of studies.

Ready to run a Meta-Analysis: Subgroup (Q-between) on your own data?

Guided setup, automatic assumption checks, effect sizes, figures and an APA write-up.

Run this test →