Methods

Meta-analysis

Meta-analysis statistically combines estimates from separate studies when their questions, designs, and outcomes can support a meaningful synthesis. Common methods use weighted averages. The pooled number is an answer to a specified model-based question, not an automatic verdict that overrides bias or differences between studies.

Can compatible study estimates be combined, and what does the combined estimate mean given their differences?

How it works

  1. Start with a justified review and define the comparison, effect metric, time point, and handling of dependent results. Decide whether the studies address a sufficiently coherent question before choosing to pool.
  2. Choose and justify the synthesis model and estimate uncertainty. Examine study-level risk of bias and heterogeneity; distinguish uncertainty about an average effect from variation in effects across settings.
  3. Present individual estimates alongside the pooled result and conduct justified sensitivity analyses. Explain how model choices, influential studies, and missing evidence affect the interpretation rather than reporting only the summary diamond.

Evidence: (Cochrane Handbook, Chapter 10: Analysing Data and Undertaking Meta-Analyses, n.d.); (Cochrane Handbook, Chapter 4: Searching for and Selecting Studies, n.d.)

Pooling delayed-recall comparisons

Teaching scenario

Synthetic example: A methods class assembles hypothetical trials comparing study routines. Some measure recall after a day, others after a month, and several use different score scales. Before computing anything, learners decide which time points and constructs can meaningfully share a synthesis and how each study will contribute only the intended independent evidence. They sketch a forest plot with labeled placeholders, then discuss how one unusually influential study might change the answer. No pooled effect or actual review outcome is claimed. The exercise is about specifying an estimand before producing a persuasive graphic.

This example is constructed for learning. It does not report an observed research result.

What can you conclude?

Can improve precision and examine consistency across studies when synthesis assumptions are appropriate.

Evidence: (Cochrane Handbook, Chapter 10: Analysing Data and Undertaking Meta-Analyses, n.d.); (Cochrane Handbook, Chapter 4: Searching for and Selecting Studies, n.d.)

Limits and common mistakes

  • Pooling cannot repair biased primary studies or guarantee that missing studies would agree.
  • An average can conceal meaningful differences; random-effects modeling does not make incompatible studies interchangeable.

A mistake to catch

Treating a pooled effect as a universal effect, or selecting a model only after checking which gives significance.

Evidence: (Cochrane Handbook, Chapter 10: Analysing Data and Undertaking Meta-Analyses, n.d.); (Cochrane Handbook, Chapter 4: Searching for and Selecting Studies, n.d.)

Sources and inspection notes

Links checked October 21, 2026. Inspection scope is stated for each source; an access date is not a publication date.

  1. Cochrane handbook, chapter 4: Searching for and selecting studies. (n.d.). https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-04

    Sections 4.2.2, 4.3, 4.6; searching, study/report distinction, selection Official current methods. Supports reproducible multisource searches and linking multiple reports of one study. Opened and checked 21 October 2026.

  2. Cochrane handbook, chapter 10: Analysing data and undertaking meta-analyses. (n.d.). https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10

    Sections 10.1–10.3, 10.10, 10.14 Current official chapter retrieved as version 6.5, updated November 2024. Supports weighted pooling, heterogeneity, model distinctions, and sensitivity analysis; not automatic pooling. Opened and checked 21 October 2026.

Keep reading

Spielman, R. M., Jenkins, W. J., & Lovett, M. D. (2020). Psychology 2e. OpenStax. https://openstax.org/books/psychology-2e/pages/2-3-analyzing-findings

Nearest related OpenStax Psychology 2e reading. This is a course connection, not evidence for every specialist detail above.

Specialist reading: Cochrane handbook, chapter 10: Analysing data and undertaking meta-analyses. (n.d.). https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10

Related methods

Try the methods · Read a study · Explore psychological measures