What is a Pooled Variance Calculator?
A pooled variance calculator combines the variances of two or more independent samples into one estimate of the population variance they share. The pooled variance formula is sₚ² = Σ(nᵢ − 1)sᵢ² / Σ(nᵢ − 1), so each group is weighted by its degrees of freedom. For groups of 8, 10 and 9 with SDs 2.5, 3.1 and 2.8, sₚ² = 8.04. It is the variance in the pooled two-sample t-test and the mean square within groups (MSW) in one-way ANOVA.
What the Pooled Variance Calculator does
Enter each group’s size with its standard deviation or its variance, or paste raw data. The calculator returns the pooled variance sₚ² (the pooled sample variance, read “s pooled squared”), the pooled SD sₚ, the within-group sum of squares, each group’s weight, Hartley’s F-max ratio as a quick check of the equal-variance assumption, and a step-by-step solution with your numbers.
It works for 2 groups or up to 20 groups from summary statistics (10 groups with raw data). With two groups it also gives the standard error of the difference between means and the critical t, so it doubles as a pooled variance calculator for the t-test. With raw data it shows how sₚ² fits into the ANOVA sum-of-squares decomposition.
Pooled statistics such as sₚ² are worthwhile because every group carries information about the same σ². Combining the groups gives an estimate with N − k degrees of freedom instead of only nᵢ − 1, which makes t-tests and F-tests more powerful, provided the groups really do share one variance. If you need the square root on its own, use the pooled standard deviation calculator.
How to use the Pooled Variance Calculator
- Under Data input, choose n and SD, n and variance, or Raw data.
- Fill in one row per group (or one box per group for raw data). Use Add group for more groups; each group needs n ≥ 2.
- Read the pooled variance and pooled SD at the top of the results. Check the notes: a large F-max ratio means pooling may not be appropriate.
- Open the pooling table to see each group’s weight, and the step-by-step solution for the link to the t-test or ANOVA.
How to calculate pooled variance
Calculating pooled variance by hand needs only each group’s sample size nᵢ and its standard deviation sᵢ or variance sᵢ². Follow four steps:
- Get each group’s variance. If you have standard deviations, square them: sᵢ² = sᵢ × sᵢ.
- Multiply each variance by nᵢ − 1, the group’s degrees of freedom. Each product is that group’s sum of squared deviations from its own mean.
- Add the products to get the within-group sum of squares, SSW = Σ(nᵢ − 1)sᵢ².
- Divide by the total degrees of freedom, Σ(nᵢ − 1) = N − k. The result is the pooled variance sₚ². Take its square root if you also need the pooled SD.
Pooled variance formula
| Symbol | Meaning |
|---|---|
| sₚ² | pooled variance (“s pooled squared”) |
| sᵢ² | sample variance of group i |
| nᵢ | size of group i |
| N, k | total sample size, number of groups |
| SSW | within-group sum of squares, Σ(nᵢ − 1)sᵢ² |
(nᵢ − 1)sᵢ² is exactly the sum of squared deviations of group i around its own mean, so the pooled variance is the total within-group sum of squares divided by its degrees of freedom. That is why it equals MSW in ANOVA.
The pooled variance symbol is sₚ² (also written s²ₚ, s_p² or S²pooled). Some textbooks write the two-group pooled variance equation with denominator n₁ + n₂ − 2; that is N − k with k = 2.
Worked example: pooled variance of three groups
Three fertilizer plots, from sizes and SDs
Problem. Plant heights (cm) from three fertilizers give n = 8, s = 2.5; n = 10, s = 3.1; n = 9, s = 2.8. Find the pooled variance that a one-way ANOVA would use.
- 01Variances: 2.5² = 6.25, 3.1² = 9.61, 2.8² = 7.84.
- 02Weighted: 7 × 6.25 = 43.75, 9 × 9.61 = 86.49, 8 × 7.84 = 62.72.
- 03SSW = 43.75 + 86.49 + 62.72 = 192.96.
- 04df = 7 + 9 + 8 = 24 (= 27 − 3).
- 05sₚ² = 192.96 / 24 = 8.04 cm², so sₚ = √8.04 = 2.8355 cm.
Answer. The pooled variance is 8.04 cm². In the ANOVA table this is MSW with 24 degrees of freedom. The groups’ weights are 29.2%, 37.5% and 33.3%, and the F-max ratio is 9.61 / 6.25 = 1.54, so pooling is reasonable.
Worked example: pooled variance for a two-sample t-test
Two classes take the same exam. Class A has n₁ = 12, mean 75.2 and s₁ = 8.1; class B has n₂ = 15, mean 68.9 and s₂ = 7.4. Using the two-group pooled variance equation, sₚ² = (11 × 8.1² + 14 × 7.4²) / (12 + 15 − 2) = (721.71 + 766.64) / 25 = 1488.35 / 25 = 59.534.
The pooled t-test standard error is √[59.534 × (1/12 + 1/15)] = 2.9883, so t = (75.2 − 68.9) / 2.9883 = 2.108 with 25 degrees of freedom. That is just above the two-tailed 5% critical value of 2.060, so the difference in means is significant at α = 0.05. The variance ratio is 65.61 / 54.76 = 1.20, well under 4, so pooling is fine.
| Class | nᵢ | sᵢ | sᵢ² | nᵢ − 1 | (nᵢ − 1)sᵢ² |
|---|---|---|---|---|---|
| A | 12 | 8.1 | 65.61 | 11 | 721.71 |
| B | 15 | 7.4 | 54.76 | 14 | 766.64 |
| Σ | 27 | 25 | 1488.35 |
Pooled variance vs the variance of all data combined
Merging every group into one list and computing a single variance gives a different, usually larger, number. That combined variance measures spread around the grand mean, so it includes the differences between group means. The pooled variance measures spread around each group’s own mean only.
In the raw-data example on this page the group means are 15.5, 21.2 and 14.0. The pooled variance is 60.3 / 15 = 4.02, but the variance of all 18 values merged is 12.9706, because the gap between the means adds to the spread. Use the pooled variance for t-tests and ANOVA; use the merged variance only to describe all observations as one sample.
| Pooled variance, sₚ² | Combined (merged) variance | |
|---|---|---|
| Spread measured around | each group’s own mean | the grand mean of all values |
| Includes differences between means | No | Yes |
| Degrees of freedom | N − k | N − 1 |
| Raw-data example on this page | 4.02 | 12.9706 |
| Used for | pooled t-test, ANOVA, effect sizes | describing all values as one sample |
When is it safe to pool variances?
Pooling is only justified if the populations have the same variance. Formal tests exist (the F-test for two groups; Levene’s test or Bartlett’s test for more), but they have low power in small samples and Bartlett’s test is very sensitive to non-normality. Many analysts use simple rules of thumb instead:
- Largest variance ÷ smallest variance (Hartley’s F-max) below about 4, equivalent to an SD ratio below 2. The calculator reports this ratio for you; the table below shows how to read it.
- Group sizes roughly equal. With balanced designs the pooled t-test and ANOVA are fairly robust to unequal variances; with unbalanced designs they are not.
- If the rules fail, use Welch’s t-test for two groups or Welch’s ANOVA for several.
| Largest variance ÷ smallest (F-max) | What it suggests |
|---|---|
| Under 2 | Variances are similar; pooling is safe. |
| 2 to 4 | Pooling is usually fine, especially with equal group sizes. |
| Over 4 (SD ratio over 2) | Equal variances are doubtful; use Welch’s t-test or Welch’s ANOVA. |
Pooled variance in Excel, R and Python
Few tools have a built-in pooled variance function, but the formula for pooled variance fits on one line:
- Excel or Google Sheets: =SUMPRODUCT(B2:B4-1,C2:C4^2)/(SUM(B2:B4)-ROWS(B2:B4)), with sizes in column B and SDs in column C. Use C2:C4 instead of C2:C4^2 if column C already holds variances.
- R: sum((n - 1) * s^2) / sum(n - 1) for vectors n and s. t.test(x, y, var.equal = TRUE) and aov() use the same pooled variance internally.
- Python (NumPy): np.sum((n - 1) * s**2) / np.sum(n - 1) for arrays n and s.
References
The formulas and rules of thumb on this page follow these sources:
- NIST/SEMATECH e-Handbook of Statistical Methods, section 1.3.5.3, Two-Sample t-Test for Equal Means (pooled variance and standard error).
- Hartley, H. O. (1950). The maximum F-ratio as a short-cut test for heterogeneity of variance. Biometrika, 37(3–4), 308–312.
- Welch, B. L. (1947). The generalization of “Student’s” problem when several different population variances are involved. Biometrika, 34(1–2), 28–35. The alternative when variances differ.
Common use cases
- Pooled two-sample t-test: sₚ² feeds the standard error of the difference; see the Two-Sample t-Test Calculator and Independent Samples t-Test: Formula, Assumptions and Worked Example.
- One-way ANOVA: sₚ² is the mean square within groups, the denominator of F; see the ANOVA Calculator and One-Way ANOVA: Definition, Formulas and a Worked Example by Hand.
- Effect size: the square root of the pooled variance is the denominator of Cohen’s d; see the Effect Size Calculator.
- Repeatability studies: pool the variances of several batches or operators to estimate measurement precision.
- Sample-size planning: use a pooled variance from earlier studies as the planning value of σ².
Assumptions and limitations
- Assumes equal population variances. Pooling very different variances understates uncertainty for the more variable group and overstates it for the other.
- Groups must be independent, and each must have at least two observations (nᵢ − 1 = 0 gives no variance).
- The pooled variance is not the variance of all observations merged; that larger number also includes between-group differences (see the decomposition table in raw-data mode).
- Like any variance, it is sensitive to outliers within groups, and a large group’s outlier carries a large weight.
Frequently asked questions
What is pooled variance?
How do I calculate pooled variance?
What is the formula for the pooled variance sₚ²?
What is pooled variance used for?
Is pooled variance the same as MSW in ANOVA?
Why is pooled variance weighted by n − 1?
What if my groups have very different variances?
How do I calculate a pooled proportion?
Cite this page
Statculator. (2026, November 14). Pooled Variance Calculator. Retrieved November 14, 2026, from https://statculator.com/descriptive-statistics/pooled-variance-calculator/