When two independent samples share a common variance, their pooled estimate combines information from both groups. Each sample contributes its size minus one degree of freedom, giving a total of n1 + n2 − 2. That count helps set the t statistic’s uncertainty and standard error.

Multiple Choice

In the two-sample t-test with equal variances, what are the degrees of freedom used for the pooled variance?

In a two-sample t-test with equal variances, you pool the two sample variances to get a single estimate of the common variance. The degrees of freedom for that pooled estimate come from how much information you have to estimate that variance from both samples. Each sample contributes its own df: (n1 − 1) from the first sample and (n2 − 1) from the second. Add them together and you get (n1 − 1) + (n2 − 1) = n1 + n2 − 2. This pooled variance, with those degrees of freedom, is then used in the denominator of the t statistic along with the factor sqrt(1/n1 + 1/n2). Thus the appropriate degrees of freedom are n1 + n2 − 2. Using n1 + n2 would ignore the two estimated means, n1 + n2 − 1 would correspond to a single combined sample rather than two separate samples, and n1 − 1 or n2 − 1 would reflect only one sample’s variance, not the pooled estimate from both.

When two groups are compared with a pooled two-sample t method, their sample sizes help determine how much information supports the estimate of shared variation. The degrees of freedom for that pooled variance are (n_1+n_2-2). The formula may look like a small bookkeeping detail, but it reflects a simple idea: each sample uses one piece of information to estimate its own mean.

Why does the formula subtract two?

Suppose one sample has (n_1) observations and the other has (n_2). To estimate how much the observations vary within each group, we measure their distances from that group’s sample mean. But the mean is calculated from the same observations, so those distances are not all free to vary independently.

For each sample, estimating its mean costs one degree of freedom. That leaves:

  • (n_1-1) degrees of freedom for the first sample’s variance

  • (n_2-1) degrees of freedom for the second sample’s variance

The pooled variance combines information from both groups. So the degrees of freedom add:

[

(n_1-1)+(n_2-1)=n_1+n_2-2

]

That’s the whole story behind the minus two: one estimated mean from each group.

How pooling works

The pooled variance is a weighted combination of the two sample variances. It gives more weight to the group with more observations, since a larger sample generally provides more information about its group’s spread.

The formula is:

[

s_p^2 =

\frac{(n_1-1)s_1^2+(n_2-1)s_2^2}

{n_1+n_2-2}

]

Here, (s_1^2) and (s_2^2) are the sample variances, while (s_p^2) is the pooled estimate. The weights, (n_1-1) and (n_2-1), are the degrees of freedom contributed by each sample. The denominator is their total.

This is not just a way to average two numbers. A plain average would give a small group the same influence as a large one. Pooling instead reflects how much data each variance estimate is based on.

The pooled standard deviation is the square root of the pooled variance:

[

s_p=\sqrt{s_p^2}

]

It enters the standard error of the difference between the two sample means:

[

SE=s_p\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}

]

The standard error estimates how much the difference between sample means might vary from one sample to another. In everyday terms, it tells us how much wiggle room to expect around that observed difference.

A quick example

Imagine comparing sales figures from two branches. The first branch has 12 observations, and the second has 10. Their degrees of freedom for the pooled variance are:

[

12+10-2=20

]

You can also see this by adding each group’s contribution:

[

(12-1)+(10-1)=11+9=20

]

Those 20 degrees of freedom describe the information used to estimate the common variance. They do not mean there are 20 observations; there are 22 in total. The difference comes from estimating two separate group means.

Why assume equal variances?

Pooling rests on an important assumption: the two populations have the same variance, even if their sample variances differ a little. In a business setting, that might mean treating the spread in order values at two branches as broadly similar, or assuming two production lines have comparable variation.

The assumption doesn’t say the groups have the same average. Their means can differ; that’s often the point of the comparison. It says only that their underlying spreads are close enough to treat as one shared variance.

When that assumption is reasonable, pooling can give a useful estimate because it draws on information from both samples. But if one group’s spread is much larger than the other’s, forcing a shared variance can distort the standard error. The calculation may still produce a number, but that number can give a misleading picture of uncertainty.

A common alternative is Welch’s two-sample method. It does not assume equal variances, and its degrees of freedom are calculated differently—often resulting in a non-integer value. That may seem less tidy, but it can be a better fit when the groups have noticeably different spreads or sample sizes.

Why not subtract one?

It can be tempting to subtract only one because a single sample variance uses (n-1) degrees of freedom. But there are two samples here, and each has its own mean. Each estimated mean reduces the available information for measuring spread within its group.

Another way to picture it: imagine two small teams each measuring the time needed to complete a task. Each team’s observations are compared with its own average time, not with one shared average. That means each team has its own constraint. Combining their variance information doesn’t erase either constraint; both remain.

This is also why (n_1+n_2-1) doesn’t fit the pooled calculation. That expression can arise when data are treated as one combined sample with one overall mean. The pooled two-sample method, by contrast, keeps the two group means separate.

And using only (n_1-1) would count the first sample’s information while ignoring the second sample’s contribution. The pooled estimate depends on both groups, so its degrees of freedom must account for both.

A small formula, a useful idea

The rule is compact:

[

\boxed{\text{Pooled-variance degrees of freedom}=n_1+n_2-2}

]

Its meaning is more memorable than the formula itself. Each sample contributes its size minus one, because each sample mean has been estimated. Add those contributions, and you have the degrees of freedom for the pooled variance.

That same way of thinking helps with many statistical formulas. Rather than memorizing every minus sign, ask what has been estimated and how many groups contribute information. Here, two group means are estimated, so two degrees of freedom are used. The pooled variance then brings the remaining information together to describe the shared spread.