Statistics · BEGINNER LESSON
Effect size
Effect size is a numerical description of how large a research result is, such as the size of a difference between groups or the strength of an association. It is reported with an estimate of uncertainty so readers can judge how precise that size is.
What it is
Effect size is the amount of the quantity of interest in a study: for example, the difference between group means, a correlation, or another measure that captures the magnitude of a phenomenon. Unlike a yes/no significance test, it tells how large the result is on a continuous scale. Researchers often report effect size together with a measure of uncertainty such as a standard error or confidence interval (CI). CI means confidence interval, a range of plausible population values for the estimate. [0,2]
[1] Effect sizes for experimental research - PMC - NIH [3] Inclusion of Effect Size Measures and Clinical Relevance in ... - PMC
How it works
Some effect sizes are unstandardized, so they stay in the original units of the outcome. Others are standardized, meaning they are divided by a measure of spread such as the standard deviation (SD), which helps compare studies that use different measurement tools. In continuous outcomes, one common standardized form is Cohen’s d, also called the standardized mean difference; it is based on two means and a standard deviation. Another family of standardized measures includes f^2, which compares variation between group means with variation within groups. [0,1,2]
[1] Effect sizes for experimental research - PMC - NIH [2] Estimating the Size of Treatment Effects: Moving Beyond P Values [3] Inclusion of Effect Size Measures and Clinical Relevance in ... - PMC
Why researchers report it
Effect sizes help describe the direction and magnitude of an effect, and they can be more informative than a P value alone. They can also support comparison across studies and later synthesis in systematic reviews and meta-analyses. In some designs, the effect of interest is conditional on other independent variables, so the chosen effect size should match that adjusted meaning. [0,1,2]
[1] Effect sizes for experimental research - PMC - NIH [2] Estimating the Size of Treatment Effects: Moving Beyond P Values [3] Inclusion of Effect Size Measures and Clinical Relevance in ... - PMC
Limitations
Effect sizes are not automatically comparable across all studies. Their meaning depends on study design, outcome reliability, random error, and whether the studies being compared are reasonably similar. Standardized effect sizes are dimensionless, so they do not have an intrinsic interpretation by themselves; their practical meaning still depends on context and the uncertainty around the estimate. [0,1,2]
[1] Effect sizes for experimental research - PMC - NIH [2] Estimating the Size of Treatment Effects: Moving Beyond P Values [3] Inclusion of Effect Size Measures and Clinical Relevance in ... - PMC
SEE THE IDEA
How effect size is built and used
Measure what happened in each group or condition.
Summarize the size of the difference, association, or treatment effect.
Optionally divide by spread such as SD to make results easier to compare.
Add CI or standard error to show precision.
Judge magnitude together with design and context.
A study outcome is summarized as an effect size, optionally standardized, then interpreted with uncertainty and study context.
Effect size is not the whole conclusion; it is one numeric summary that must be read with its uncertainty and design features.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.