What Is Pooled Standard Deviation?
The pooled standard deviation (Sp) is a weighted average of two sample standard deviations, combined into a single estimate of the common population standard deviation. It is used when you assume two independent samples come from populations that share the same variance. The pooled estimate is central to the two-sample t-test, Cohen's d effect size, and confidence intervals for the difference between two means.
How to Use This Calculator
Enter the size of each sample (\(n_1\) and \(n_2\)) and each sample's standard deviation (\(s_1\) and \(s_2\)). The calculator returns the pooled standard deviation, the pooled variance (\(S_p^2\)), and the degrees of freedom (\(n_1 + n_2 - 2\)). Each sample must contain at least two observations so the degrees of freedom stay positive.
The Formula Explained
The pooled variance weights each sample's variance by its degrees of freedom (\(n - 1\)):
$$S_p^{2} = \dfrac{(n_1 - 1)\,s_1^{2} + (n_2 - 1)\,s_2^{2}}{n_1 + n_2 - 2}$$
Taking the square root gives the pooled standard deviation, \(S_p\). Larger samples contribute more heavily to the pooled estimate, which is why the variances are weighted rather than simply averaged.
Worked Example
Suppose sample 1 has \(n_1 = 10\) and \(s_1 = 5\), and sample 2 has \(n_2 = 12\) and \(s_2 = 6\). Then \((10-1)\cdot 25 = 225\) and \((12-1)\cdot 36 = 396\), summing to 621. The degrees of freedom are \(10 + 12 - 2 = 20\), so $$S_p^{2} = \frac{621}{20} = 31.05 \quad\text{and}\quad S_p = \sqrt{31.05} \approx 5.5722.$$
How Pooled SD Changes Across Scenarios
The pooled variance is a weighted average of the two sample variances, with weights equal to each sample's degrees of freedom: \(S_p^2 = \dfrac{(n_1-1)s_1^2 + (n_2-1)s_2^2}{n_1+n_2-2}\). The table applies it to four scenarios.
| Scenario | n₁ | s₁ | n₂ | s₂ | Pooled variance | Sp | df |
|---|---|---|---|---|---|---|---|
| A. Equal n, different SDs | 20 | 4 | 20 | 8 | (19·16 + 19·64)/38 = 40 | 6.3246 | 38 |
| B. Unequal n, larger sample dominates | 10 | 5 | 40 | 10 | (9·25 + 39·100)/48 = 85.9375 | 9.2702 | 48 |
| C. Identical SDs | 12 | 6 | 30 | 6 | (11·36 + 29·36)/40 = 36 | 6 | 40 |
| D. Small sample with larger SD | 8 | 12 | 50 | 4 | (7·144 + 49·16)/56 = 32 | 5.6569 | 56 |
- Scenario A: with equal sample sizes, Sp is the root-mean-square of the two SDs: \(\sqrt{(4^2+8^2)/2} = \sqrt{40} \approx 6.3246\). This is larger than the simple average of 6.
- Scenario B: the second sample has four times as many observations, so Sp (9.27) lands much closer to s₂ = 10 than to s₁ = 5.
- Scenario C: when s₁ = s₂, the weights do not matter and Sp equals that common SD, regardless of sample sizes.
- Scenario D: the 50-observation sample has the smaller SD, so Sp (5.66) is pulled toward 4 even though the small sample has a much larger SD of 12.
Interpreting the Pooled Standard Deviation
- Always between s₁ and s₂: because Sp² is a weighted average of s₁² and s₂² with positive weights, \(\min(s_1, s_2) \le S_p \le \max(s_1, s_2)\). It equals both only when s₁ = s₂.
- Estimate of a common SD: under the assumption that both populations have the same variance σ², Sp² is the combined estimate of σ². It uses information from both samples, so it is more stable than either s alone.
- Standard error of the difference in means: $$SE = S_p\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}$$ In Scenario A, \(SE = \sqrt{40}\cdot\sqrt{0.1} = \sqrt{4} = 2\).
- Cohen's d: $$d = \frac{\bar{x}_1 - \bar{x}_2}{S_p}$$ In Scenario A, a mean difference of 5 gives \(d = 5/6.3246 \approx 0.79\).
- Degrees of freedom: \(df = n_1 + n_2 - 2\) is the df of the t-distribution in the pooled (Student's) two-sample t-test. Scenario A has df = 38 and Scenario B has df = 48.
- Checking the equal-variance assumption: a common rule of thumb is that if the larger SD is more than about 2 times the smaller SD (a variance ratio above about 4), pooling is questionable. This is especially true when the sample sizes differ. In that case Welch's t-test, which does not pool variances, is typically used instead. Scenario A has an SD ratio of exactly 2, and Scenario D has a ratio of 3.
Conventions and edge cases
Each case below was run through this calculator; the result column is what it actually returns.
| Case | Input | Result |
|---|---|---|
| Groups of 10 and 12 (SD 5 and 6) | 10, 5, 12, 6 | 5.5723 |
| One value per group | 1, 5, 1, 6 | The pooled standard deviation needs at least 2 values in each group. |
Checked on 2026-11-08
FAQ
When should I pool standard deviations? Pool when the two groups are assumed to have equal population variances. If variances differ greatly, use Welch's t-test instead.
Why use n - 1 instead of n? Using \(n - 1\) (Bessel's correction) gives an unbiased estimate of variance from a sample.
Does sample order matter? No. Swapping the two samples produces the same pooled standard deviation.