Choose your reading route
eduKateSG · Why Mathematics? · Pooled Standard Deviation, Degrees of Freedom and Common Spread
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 5–9
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 10–15
Check assumptions
See where design and calibration matter.Route 4 · Chapters 16–20
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 21–25
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important when several groups share a common spread? A pooled standard deviation combines within-group variance using degrees-of-freedom weights. It supports comparisons of means when a common-variance model is defensible.
Pooling is not simply averaging two standard deviations. Variances combine before the square root, and larger groups contribute more degrees of freedom.
Did you know? Two groups can have very different means yet still be pooled for within-group spread, because pooling removes each group's own mean before combining residual variation.
Full chapter index
Chapters 1–4 · Understand the question
Chapters 5–9 · Reproduce the example
Chapters 10–15 · Check assumptions
Chapters 16–20 · Explore applications
Chapters 21–25 · Study and verify
CHAPTER 1 OF 25 · Understand the question
1. The common-within-group-spread question
Suppose two classes use the same measurement scale but have different average scores. We may want one estimate of typical within-class variability.
A pooled standard deviation combines the squared residual information from both groups while keeping between-group mean differences out of the spread estimate.
For independent samples with sizes n1 and n2 and standard deviations s1 and s2, NIST gives pooled s as the square root of [((n1−1)s1²+(n2−1)s2²)/(n1+n2−2)].
Read the NIST two-process comparison. NIST also states that equivalence of standard deviations should be examined before assuming it.
Group A has n1=10 and s1=4. Group B has n2=15 and s2=6. Their variance contributions are 9×16=144 and 14×36=504.
Total squared-residual contribution is 648 and total degrees of freedom is 23. Pooled variance is 648/23=28.1739, so pooled standard deviation is about 5.308.
| Group | n | s | df times variance |
|---|---|---|---|
| A | 10 | 4 | 144 |
| B | 15 | 6 | 504 |
Each sample variance uses n−1 degrees of freedom because its group mean was estimated from the same data.
The contribution (n−1)s² reconstructs that group's residual sum of squares, allowing honest addition across groups.
| Residual sum | Residual df | Pooled SD |
|---|---|---|
| 648 | 23 | sqrt(648/23) = 5.308 |
The simple average of 4 and 6 is 5, which ignores unequal group sizes and combines on the wrong mathematical scale.
Even a sample-size-weighted average of standard deviations is generally not the pooled standard deviation because squared residuals, not distances, add.
Pooling estimates within-group spread, so each group is centred on its own mean before residuals are combined.
If the scientific question concerns total spread after merging all observations around one grand mean, a different calculation is needed and between-group differences matter.
The classical pooled two-sample t procedure assumes the populations have a common variance. Similar sample standard deviations alone do not prove that model.
Use subject knowledge, design, graphics and appropriate diagnostics; avoid a mechanical pre-test as the only justification.
If common variance is doubtful, Welch's two-sample method uses separate variance terms and adjusted degrees of freedom.
NIST presents the unequal-variance formula beside the pooled case, making the modelling choice explicit.
For k independent groups under a common-variance model, add each residual sum of squares (n_j−1)s_j² and divide by total residual degrees of freedom sum(n_j−1).
This is the within-group mean square behind one-way analysis of variance.
Repeated measurements across instruments, wafers or days may be pooled when the design seeks one repeatability component and conditions are sufficiently comparable.
NIST provides measurement examples of pooled standard deviations across configurations. See a NIST laboratory example.
The ordinary formula assumes independent residual information. Paired measurements, repeated readings and nested groups require a model that preserves their dependence.
Counting correlated repeats as independent inflates apparent degrees of freedom.
Squaring residuals makes pooled standard deviation sensitive to outliers. Strong skew or heavy tails can also weaken procedures built on normal common-variance assumptions.
Inspect group distributions separately before compressing them into one number.
Group variances are in squared units; the pooled variance remains squared; the final square root restores the measurement unit.
All groups must use compatible scales and units before pooling.
Use the actual valid sample size and standard deviation for each group after applying a declared missing-data rule.
Do not keep the original n if some observations were excluded from the variance calculation.
Square unrounded standard deviations when possible. Early rounding of s1 and s2 is magnified by squaring and can affect the final result.
Round the pooled standard deviation at the end to a precision supported by the data.
Create columns for n, degrees of freedom, s, s² and (n−1)s². Sum the last and degrees-of-freedom columns, divide and take the square root.
Add a check that every group uses the same unit and definition.
Accept group sizes and variances, reject n below two, sum residual degrees of freedom and return both pooled variance and pooled standard deviation.
Test identical spreads, unequal sizes and the worked example value 5.308.
Some standardised mean differences divide a mean difference by a pooled standard deviation. That expresses the difference in common within-group spread units.
Small-sample corrections and interpretation belong to the specific effect-size method; pooled s alone does not prove importance.
One pooled number can conceal a group with much larger variability, a data error or a changing process. Always display group-specific spreads too.
Pooling should summarise a defensible common pattern, not erase heterogeneity.
Pooled standard deviations appear in experiments, manufacturing, educational research, clinical studies and measurement science. Analysts use them to estimate common repeatability or standardise differences.
The transferable skill is deciding whether groups are comparable enough to share one variance parameter.
The weighted-variance guide explains weighted squared deviations. The root-mean-square guide clarifies why squares combine before the square root.
Students should reconstruct each residual sum of squares, add degrees of freedom and state the common-variance assumption. Parents can ask whether the groups used the same measurement process and whether separate spreads were inspected.
The calculation is strongest when accompanied by a graph and a limitation sentence.
- Pool n=8, s=3 with n=12, s=5.
- Compare the pooled result with the simple average of 3 and 5.
- Explain why very different group means do not enter the within-group formula.
### Is pooled standard deviation always between the group standard deviations? With positive residual degrees-of-freedom weights in the standard two-group formula, the pooled variance is a weighted average of group variances, so the pooled standard deviation lies between them.
### Can I pool groups with different units? No. Convert to one defensible common scale first or keep them separate.
### Must sample sizes be equal? No. Degrees-of-freedom weighting handles unequal sizes, provided the model assumptions are reasonable.
Continue through the Mathematics Learning Hub and compare NIST's equal-variance and Welch formulas using the same two samples.
Extended practice laboratory
Practice case 1: List group sizes
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 1. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 2: Convert sizes to degrees of freedom
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 2. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 3: Square each group SD
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 3. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 4: Recover residual sums of squares
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 4. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 5: Add residual sums
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 5. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 6: Add degrees of freedom
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 6. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 7: Divide to pooled variance
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 7. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 8: Take the square root
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 8. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 9: Compare simple average
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 9. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 10: Change one group size
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 10. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 11: Change one group spread
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 11. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 12: Keep group means separate
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 12. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 13: Inspect equal-variance assumption
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 13. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 14: Use Welch alternative
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 14. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 15: Add a third group
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 15. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 16: Check independence
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 16. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 17: Inspect outliers
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 17. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 18: Check common units
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 18. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 19: Handle missing observations
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 19. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 20: Display group spreads
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 20. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 21: Connect to effect size
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 21. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 22: Audit laboratory repeatability
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 22. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 23: Avoid homogeneity overclaim
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 23. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 24: Write a pooled-spread conclusion
Setup: Use the article’s pooled-standard-deviation mechanism and change one declared input for case 24. Task: Recalculate from raw values, preserve units and state which mathematical or design condition changed.
Check: Check the relevant invariant, separate the number from its claim, and name the next diagnostic or source of evidence. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.