Paired t-test

A paired t-test is a hypothesis test for two related measurements, like the same people before and after treatment. In Intro to Statistics, you use it to test whether the average difference between pairs is different from zero.

What is the Paired t-test?

A paired t-test is the test you use in Intro to Statistics when the two data sets are linked, not independent. Instead of comparing two separate sample means, you calculate one difference for each pair and test the mean of those differences.

That setup matters because the pairing removes a lot of random variation. If you measure the same students before and after a study session, each student acts as their own comparison point. The test looks at whether the average change across all those students is bigger than what you would expect from ordinary chance.

The key idea is that the raw data are converted into difference scores first. For each pair, you subtract one measurement from the other, then analyze those differences as a single sample. If the differences cluster near zero, there is little evidence of a real shift. If they are mostly positive or mostly negative, the sample mean difference may be far enough from zero to matter.

The hypotheses are written about the population mean difference, not about the two raw means separately. A common null hypothesis is that the mean difference equals 0, which means the treatment, change, or matched comparison did not create a systematic shift. The alternative can be one-sided or two-sided depending on the question.

The test statistic compares the sample mean difference to the standard deviation of the differences and the size of the sample. Smaller spread in the differences gives you a more precise estimate, which makes it easier to detect a real effect. That is why paired tests are often more sensitive than independent-samples tests when the pairing is valid.

The main assumption is that the distribution of the differences is approximately normal, especially for small samples. The paired t-test is not about whether each raw measurement is normal, but whether the difference scores are reasonably bell-shaped or at least not badly skewed.

Why the Paired t-test matters in Intro to Statistics

Paired t-test shows up any time Intro to Statistics moves from separate groups to repeated or matched measurements. It is the clean way to answer questions like, did a tutoring program change test scores, did reaction time improve after practice, or did a matched pair of twins differ after one got a new diet?

This term also teaches a bigger statistical habit: match the test to the structure of the data. If you ignore the pairing and treat related values like independent samples, you can distort the standard error and get the wrong p-value. The paired t-test fixes that by focusing on within-pair change instead of between-person noise.

It also connects directly to hypothesis testing language. You are not just checking whether two averages look different. You are testing whether the average difference in the population is likely to be zero or not. That shift in wording shows up again and again in statistics problems, so getting comfortable with it helps you read questions carefully and set up the right null and alternative hypotheses.

In a class, this term often appears in practice problems with before-and-after data, matched case studies, or repeated measures tables. If you can spot the pairing, define the differences, and interpret the result in context, you are doing the exact move the course is aiming for.

How the Paired t-test connects across the course

Matched Samples

Matched samples are the data structure that makes a paired t-test possible. Each observation has a natural partner, so the analysis focuses on the difference within each pair instead of comparing two unrelated group averages. If the pairing is weak or artificial, the test is not the right choice.

Dependent Samples

Dependent samples means the values are connected in some way, often because they come from the same person twice or from specially matched pairs. A paired t-test is designed for dependent samples, while independent-samples methods assume the observations do not influence each other.

Standard Deviation of Differences

This is the spread measure that matters in a paired t-test. The test statistic uses the variability of the difference scores, not the variability of each original group separately. Smaller spread in differences gives more precise evidence about whether the mean change is really different from zero.

Null Hypothesis

For a paired t-test, the null hypothesis usually says the population mean difference is 0. That means there is no average change across the matched pairs. Your whole test is built around checking whether the sample differences are far enough from 0 to reject that claim.

Is the Paired t-test on the Intro to Statistics exam?

A problem set or quiz item will usually give you matched data, then ask you to choose the correct test, state hypotheses, or interpret a p-value. Your first move is to check whether the same people were measured twice or whether the pairs are naturally linked. If they are, you use the paired t-test and work with the difference scores.

For calculations, you may be asked to find the mean difference, the standard deviation of the differences, and the t statistic. On a written answer, you should say what the test is comparing in context, such as average before-and-after change, not just "the means." If the p-value is small, you explain that the data give evidence that the average difference is not zero.

A common mistake is trying to use an independent-samples t-test just because there are two sets of numbers. If the values come in pairs, that shortcut ignores the design and weakens the analysis.

The Paired t-test vs Independent t-test

A paired t-test uses related observations, while an independent t-test uses two separate groups with no natural pairing. The paired version tests the mean of difference scores, but the independent version compares two group means directly. If the same people are measured twice, or each subject has a matched partner, you want the paired test.

Key things to remember about the Paired t-test

• A paired t-test compares two related measurements by turning each pair into one difference score.

• In Intro to Statistics, the null hypothesis usually says the population mean difference is 0.

• This test works best for before-and-after studies, repeated measures, or matched pairs.

• The spread of the differences matters, because the test statistic uses the standard deviation of differences.

• If the data are not paired, you should not use a paired t-test just because there are two samples.

Frequently asked questions about the Paired t-test

What is a paired t-test in Intro to Statistics?

A paired t-test is a hypothesis test for two related measurements, such as the same class before and after a review session. It tests whether the average difference between the pairs is different from zero. The whole method depends on the data being matched or repeated.

How do you know if you should use a paired t-test?

Use it when each value in one sample has a direct match in the other sample. Common examples are before-and-after data from the same people, twin studies, or matched case-control pairs. If the groups are separate and unrelated, you need a different test.

What is the null hypothesis for a paired t-test?

The null hypothesis usually says the mean of the population differences is 0. That means there is no average change across the matched pairs. The alternative hypothesis says the mean difference is not 0, or is greater than 0, or less than 0 depending on the question.

What is the difference between a paired t-test and an independent t-test?

A paired t-test looks at change within the same subject or matched pair, while an independent t-test compares two separate groups. The paired test uses difference scores, which often makes it more powerful when the data are actually matched. The independent test ignores pairing, so it is not the right choice for repeated measures.