Biostatistics · Intermediate

How a Meta-Analysis Works

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

A meta-analysis synthesizes individual study results by calculating an effect size for each study, which serves as the fundamental unit of analysis. These effect sizes are combined into a summary effect, weighted by their precision (often driven by sample size), to assess the overall consistency and magnitude of an effect across studies.

Intuition

Think of a meta-analysis as a weighted average of evidence, where each study contributes a piece of the puzzle. The 'effect size' is the specific measurement of interest from each study, acting as the currency of the analysis. Just as a larger sample provides a clearer picture, studies with greater precision (narrower confidence intervals) are given more weight in the final calculation. This ensures that more reliable data influences the summary effect more heavily than less precise estimates.

Derivation

  1. Identify the effect size for each individual study, which represents the magnitude of the treatment effect or relationship.
  2. Determine the precision of each study, reflected by the width of its confidence interval.
  3. Assign weights to each study based on their precision, with larger sample sizes typically driving greater precision and weight.
  4. Combine the individual effect sizes using the assigned weights to calculate a summary effect.
  5. Assess statistical significance by checking if the 95% confidence interval of the summary effect includes the null value.

Worked example

Problem

A researcher is conducting a meta-analysis on a medical intervention. Study A has a risk ratio of 0.8 with a narrow confidence interval, while Study B has a risk ratio of 0.8 with a wide confidence interval. Which study contributes more to the summary effect, and what is the null value for this ratio-based measure?

  1. Identify the measure type: Risk ratio is a ratio-based measure.
  2. Determine the null value: For ratio-based measures, the null value is 1.0.
  3. Assess precision: Study A has a narrow confidence interval, indicating high precision. Study B has a wide confidence interval, indicating low precision.
  4. Apply weighting rule: Studies with greater precision are assigned more weight.
  5. Conclusion: Study A contributes more to the summary effect because it has greater precision.

Answer Study A contributes more to the summary effect. The null value for this ratio-based measure is 1.0.

Common mistakes

  • Confusing the null value for ratio-based measures (1.0) with difference-based measures (0.0).
  • Assuming that precision is the sole determinant of study weight, ignoring other potential factors.

Animation

Animation coming soon

The animated version of this lesson is being prepared for the web. In the Signals AI app, lessons are drawn step by step.

Code

This is a text lesson: it has no animation code yet.

ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.