Cohen's d — Definition, Formula & Example
Cohen's d is a number that tells you how far apart two group means are, measured in standard deviations. It shows how big a difference is, not just whether the difference is statistically significant.
For two populations with means and and a common standard deviation , the standardized mean difference is . Cohen's is the sample estimate of . It replaces the population means with the sample means and , and it replaces with the pooled sample standard deviation . Cohen proposed , , and as rough benchmarks for small, medium, and large effects.
Key Formula
Where:
- = Sample means of group 1 and group 2
- = Sample standard deviations of the two groups
- = Sample sizes of the two groups
- = Pooled standard deviation, a weighted combination of the two groups' spreads
- = Effect size in standard deviation units (sign shows which group is larger)
Worked Example
Problem: A garden club grows 25 tomato plants with a new fertilizer and 25 without it. The fertilized plants have a mean height of 34 inches with standard deviation 6 inches. The unfertilized plants have a mean height of 30 inches with standard deviation 8 inches. Find Cohen's d and interpret it.
Step 1: Find the difference in sample means.
Step 2: Compute the pooled standard deviation. Each group has degrees of freedom, so there are 48 in total.
Step 3: Divide the mean difference by the pooled standard deviation. The inches cancel, so d has no units.
Answer: Cohen's d is about 0.57. On average, the fertilized plants are 0.57 pooled standard deviations taller, which counts as a medium effect by Cohen's benchmarks.
Why It Matters
In AP Statistics, a two-sample t-test tells you whether a difference in means is statistically significant, but not whether it is large. Researchers in psychology, education, and medicine report Cohen's d so readers can judge practical importance. Because d has no units, it also lets meta-analyses combine studies that used different scales. Researchers use it as well when choosing a sample size before running an experiment.
Common Mistakes
Mistake: Treating a small p-value as proof of a large effect, or dividing the mean difference by the standard error instead of the standard deviation.
Correction: A huge sample can make a tiny difference statistically significant. Cohen's d uses the standard deviation, which does not shrink as n grows, so it measures the size of the effect separately from sample size.
