Why can two-sample degrees of freedom be fractional?
You will be able to: Use technology-computed df without pooling group variances.
Why can two-sample degrees of freedom be fractional?
One group’s mean is estimated much less precisely than another’s. A two-sample procedure should reflect that imbalance rather than simply adding the sample sizes.
A useful starting point: How do you estimate a difference between independent averages? →
Words and symbols before equations
- Variance contribution A
- s₁²/n₁, uncertainty supplied by group 1.
- Variance contribution B
- s₂²/n₂, uncertainty supplied by group 2.
- Welch–Satterthwaite df
- An approximate reference df based on A,B and the group sizes.
- Unpooled
- No assumption that population variances are equal.
What this picture assumes
Synthetic summaries, with positive sample SDs. Random samples without replacement use source populations of N=100000 each, satisfying 10% for these sizes. A randomized experiment does not require that sampling-fraction check. Design and shape are assumptions chosen here, not conclusions that summary statistics can verify. Strong skewness/outlier selection conservatively withholds inference at all sizes; real data require examination. Groups are independent. Welch df is used with separate variance estimates; equal population variances are not assumed.
Read the picture in three steps
- Read the axes and labels first. Identify what each symbol and line represents. Read the units and fixed conditions before comparing quantities.
- 95% interval (-0.0294, 8.0294) minutes; margin 4.0294 minutes. Chosen design, sampling fraction where applicable, and shape/size assumptions support the procedure.
- Check what the picture assumes below. Use the Explore task to predict one change before moving a control.
Connect the picture to the mathematics
Welch df=(A+B)²/[A²/(n₁−1)+B²/(n₂−1)]. This is normally computed with technology. Fractional df are valid: the t distribution is defined for positive real df.
When sizes and SDs match, both groups contribute equally and df=n₁+n₂−2. When one contribution dominates, df moves closer to that group’s n−1.
The df lies between the smaller n−1 and n₁+n₂−2. Use the reported value rather than automatically rounding it or using total n−2 in an unequal-spread setting.
A worked example, step by step
For n₁=n₂=10 and s₁=s₂=6, calculate SE and df.
- A=B=36/10=3.6.
- SE=√7.2≈2.683.
- df=7.2²/[3.6²/9+3.6²/9]=18.
- Use t with df=18 for this balanced case; equality here is a property of these summaries, not a blanket pooling rule.
A fractional df is not a fraction of a student. It is a parameter for the approximate reference distribution.
Must Welch df be an integer?
Compare with an explanation
No. Technology can evaluate t at fractional df.
Predict. Change one thing. Explain.
Hold sample sizes equal and make one SD much larger. Compare the two variance contributions and the resulting df.
On narrow screens, swipe or scroll diagrams sideways to read all labels.
95% interval (-0.0294, 8.0294) minutes; margin 4.0294 minutes. Chosen design, sampling fraction where applicable, and shape/size assumptions support the procedure.
| Quantity | Value |
|---|---|
| Estimate (minutes) | 4 |
| SE (minutes) | 2 |
| Degrees of freedom | 44.51039 |
| t* | 2.01472 |
| Group 1 variance contribution (min²) | 2.56 |
| Group 2 variance contribution (min²) | 1.44 |
Synthetic summaries, with positive sample SDs. Random samples without replacement use source populations of N=100000 each, satisfying 10% for these sizes. A randomized experiment does not require that sampling-fraction check. Design and shape are assumptions chosen here, not conclusions that summary statistics can verify. Strong skewness/outlier selection conservatively withholds inference at all sizes; real data require examination. Groups are independent. Welch df is used with separate variance estimates; equal population variances are not assumed.
Explain what you noticed: Answer the investigation prompt above. State one observation and explain it using the means, standard errors, pairing, graph scales or model assumptions. Identify what the representation cannot tell you.
Apply the idea to a fresh problem Practice →Show what you understand.
Two original questions are a starting check, not proof of mastery. Explain your choice before revealing the answer.
Original written challenge
4 points · self-check · not an official AP questionFor equal sizes n₁=n₂=16 and equal positive SDs, state df and explain what happens if one SD becomes much larger.
This response is not submitted or saved. Copy it before leaving.
Compare with the answer and four-point rubric
- 1 point: With equal contributions, df=30.
- 1 point: SE uses the sum of contributions, not their difference.
- 1 point: As one variance contribution dominates, df moves toward 15.
- 1 point: Use the updated technology value and critical value for the new interval.
Accept equivalent correct methods and explanations. This is a Refresh Kid teaching rubric, not an official AP scoring guideline.
Retrieve it before you reveal it.
RECALL 1What is the lower df bound?
The smaller of n₁−1 and n₂−1.
RECALL 2What is the upper bound?
n₁+n₂−2.
RECALL 3Should you pool by default?
No; this course model uses separate variances.
Revisit these tomorrow and a week later. Try a fresh problem and explain why the method applies.
Why can two-sample degrees of freedom be fractional?
- A=s₁²/n₁; B=s₂²/n₂.
- SE=√(A+B).
- Welch df=(A+B)²/[A²/(n₁−1)+B²/(n₂−1)].
Remember: A fractional df is not a fraction of a student. It is a parameter for the approximate reference distribution.
Conditions: Synthetic summaries, with positive sample SDs. Random samples without replacement use source populations of N=100000 each, satisfying 10% for these sizes. A randomized experiment does not require that sampling-fraction check. Design and shape are assumptions chosen here, not conclusions that summary statistics can verify. Strong skewness/outlier selection conservatively withholds inference at all sizes; real data require examination. Groups are independent. Welch df is used with separate variance estimates; equal population variances are not assumed.
Refresh Kid · AP Statistics Unit 4 · Objectives 4.7.C, 4.7.D · Review edition
Framework, scope and review status
Mapped to College Board, AP Statistics CED, Topic 4.7, objectives 4.7.C, 4.7.D. Framework effective Fall 2026, checked September 17, 2026. Unit 4 includes sampling distributions of means, one-sample and paired t inference, and independent two-sample t inference; it is part of the revised five-unit course.
Examples and datasets are synthetic, independently authored teaching material. Mean inference requires a justified design and suitable shape or sample size. Paired analysis uses one sample of differences. Independent two-sample inference uses separate variance estimates and technology-computed Welch degrees of freedom. Extreme skewness and influential observations need attention even in larger samples. This model conservatively withholds inference when those warnings are selected. Conclusions are limited by random sampling and/or assignment as appropriate.
The Organic Chemistry Tutor companion title and destination were located; the full video was not reviewed. Khan Academy’s destination was checked, but its lesson content was not fully readable by the research tool. OpenStax provides optional reference reading. No provider scripts, questions or graphics were copied. Refresh Kid is not affiliated with these providers.
GitHub’s 3D website collection informed optional spatial inspection. Our original paired-data display uses self-hosted Three.js with its MIT license. Two measurement columns are connected within each labeled student lane; horizontal position is before/after, vertical position is time, and depth separates identities rather than representing a numerical variable. Rotation can separate overlapping connectors. Exact values and differences always remain in the 2D table and labeled plot. The broken-matching option is an explicit counterexample, not a legitimate alternative analysis. No autoplay; complete teaching remains available without 3D.
Independent teacher review and observation of students remain pending. Technical checks do not certify statistical accuracy, accessibility or learning effectiveness. This is a review edition.
Released AP Statistics questions and scoring guides are optional. Older exams use the earlier framework, so check alignment before selecting parts. All practice on this page is original, not official AP material.
Learn → Explore → Practice → Review is informed by the IES learning guide. This implementation has not yet been evaluated with learners.
Want to work through this with a tutor?
Bring your question about Why can two-sample degrees of freedom be fractional? Your explanation and answers remain free to access.