Two groups, two standard deviations — and one comparison to make. When you need a single estimate of variability shared by both groups, you pool their standard deviations into one number, weighted by how much data each group contributed. This pooled standard deviation calculator takes each group’s size and standard deviation and returns the pooled variance, the pooled standard deviation, and the degrees of freedom that travel with them into a t-test.
The Pooled Standard Deviation Formula
You cannot average two standard deviations directly — variances pool, standard deviations do not. The pooled variance is a degrees-of-freedom-weighted average of the two sample variances: sp² = ((n₁ − 1)s₁² + (n₂ − 1)s₂²) ÷ (n₁ + n₂ − 2), and the pooled standard deviation is its square root. The weights are n − 1, not n, because each sample variance was itself computed with n − 1 in its denominator: pooling is really combining the two sums of squared deviations and dividing by the total degrees of freedom.
Worked Example: Two Class Sections
Section 1 has n₁ = 12 students with s₁ = 4.2 points; section 2 has n₂ = 10 with s₂ = 3.6. The pooled variance is (11 × 17.64 + 9 × 12.96) ÷ 20 = (194.04 + 116.64) ÷ 20 = 310.68 ÷ 20 = 15.534, so sp = √15.534 = 3.941, on 20 degrees of freedom. Notice the pooled value sits between 3.6 and 4.2, pulled slightly toward section 1 — the bigger group with more degrees of freedom gets more say, exactly as the weighting intends.
Where the Pooled SD Gets Used
Its home is the independent two-sample t-test in its classic Student form: the standard error of the difference between means is sp × √(1/n₁ + 1/n₂), so every such t-statistic is built on the pooled value. Cohen’s d uses it too — the standardizer that turns a mean difference into an effect size is typically the pooled SD. ANOVA generalizes the same idea: the within-group mean square is a pooled variance across all groups at once. If you meet a pooled standard deviation, a group comparison is nearby.
The Assumption You Are Making
Pooling assumes the two populations share one true variance — that the groups differ (if at all) in their means, not their spreads. When the sample SDs are close (a common rule of thumb: the larger less than twice the smaller), the assumption is comfortable. When they differ wildly, pooling manufactures a fiction — an SD that describes neither group — and the honest alternative is Welch’s t-test, which keeps the variances separate and adjusts the degrees of freedom instead. Eyeball the two SDs before you pool; this calculator’s inputs make the comparison easy.
Why Not Just Average the Two SDs?
Because standard deviation is a square root, and averages do not pass through square roots. Averaging 4.2 and 3.6 gives 3.9 — suspiciously close to the correct 3.941 here only because the group sizes are similar and the SDs are close. With unequal sizes the naive average drifts: it ignores that the group with 30 observations knows its variance far better than the group with 5. Squaring, weighting by degrees of freedom, and square-rooting at the end is not pedantry; it is what makes the pooled figure an unbiased estimate of the shared variance.
Common Mistakes
Entering variances where standard deviations belong (or the reverse) is the most frequent slip — check whether your source reports s or s². Pooling groups that should not be pooled — different measurement scales, or SDs differing threefold — is the conceptual error. And a bookkeeping one: the degrees of freedom are n₁ + n₂ − 2, not n₁ + n₂; the two lost degrees are the two group means estimated along the way, and the t-table lookup depends on getting it right.
A Note on Reporting
When you publish a pooled standard deviation, report it with its ingredients — both group sizes, both group SDs, and the degrees of freedom — so a reader can verify the pooling and judge the equal-variance assumption for themselves. A bare pooled figure hides exactly the numbers that decide whether pooling was legitimate.
Related Tools
Find companion inference tools in the Statistics Calculators hub, including the Relative Standard Deviation Calculator and the Standard Deviation Calculator.
Frequently Asked Questions
How do you calculate pooled standard deviation?
Square both SDs, weight each by its degrees of freedom (n − 1), add, divide by the total degrees of freedom (n1 + n2 − 2), and take the square root. Example: groups of 12 (s = 4.2) and 10 (s = 3.6) pool to sp = 3.941.
Why pool variances instead of standard deviations?
Variance is the additive quantity — sums of squared deviations combine directly. Standard deviation is a square root of that, so it must be pooled through its square and rooted at the end.
When should I not pool?
When the groups’ spreads are clearly different — SDs differing by a factor of two or more is the usual warning line — or when group sizes and SDs are both very unequal. Use Welch’s t-test instead, which does not assume equal variances.
What degrees of freedom go with a pooled SD?
n1 + n2 − 2. Each group spends one degree of freedom estimating its own mean. The worked example: 12 + 10 − 2 = 20 df.
Is pooled SD the same as the SD of all the data combined?
No. The combined-data SD mixes within-group spread with the gap between group means, inflating it whenever the means differ. Pooled SD captures within-group spread only, which is what group comparisons need.
Can I pool more than two groups?
Yes — the formula generalizes: weight every group’s variance by its degrees of freedom and divide by the total df. That multi-group pooled variance is exactly ANOVA’s within-group mean square.
Does pooled SD assume the groups have equal sizes?
No. Unequal sizes are handled by the degrees-of-freedom weights — the larger group simply contributes more information and pulls the pooled estimate closer to its own variance.