McNemar's Test

McNemar's test is a statistical test for paired binary data, like yes/no results from the same people before and after a treatment. In Honors Statistics, you use it to see whether the proportion changed between two related measurements.

What is McNemar's Test?

McNemar's test is the test you use in Honors Statistics when you have matched or paired samples and the outcome is binary, like yes/no, pass/fail, or improved/not improved. It asks a very specific question: did the two related measurements change in a way that looks bigger than random variation?

The setup matters. You are not comparing two separate groups of people. You are comparing the same person twice, or two subjects that were deliberately paired, so each pair gives you two linked outcomes. That means the usual two-sample methods are not the right fit, because the observations are not independent in the same way.

What makes McNemar's test different is that it focuses on the pairs where the answer changed. If someone was “yes” before and “no” after, that discordant pair counts one way. If someone was “no” before and “yes” after, that counts the other way. The pairs where the answer stayed the same do not drive the test statistic the same way, because they do not show a change in direction.

That is why this test is often used for before-after measurements, treatment studies, and case-control studies. For example, if a class surveys the same students before and after a review session to see whether more of them answer a concept correctly, McNemar's test checks whether the change in correct responses is statistically significant.

Under the null hypothesis, the number of pairs that change one way should be about the same as the number that change the other way. In the common large-sample version, the test statistic follows a chi-square distribution with 1 degree of freedom. That makes McNemar's test a nonparametric method, so you do not need a normality assumption for the binary outcome. The key requirement is the pairing itself, plus a dichotomous variable.

A common mistake is trying to use a two-proportion z-test here just because there are two time points or two conditions. If the data are matched, the pairing is the whole story. McNemar's test keeps the analysis centered on changes within each pair, which is exactly what the data are telling you.

Why McNemar's Test matters in Honors Statistics

McNemar's test shows up anywhere Honors Statistics asks you to choose the right procedure for paired categorical data. It connects the big ideas of experimental design, data type, and hypothesis testing, which is a big part of doing statistics well instead of just memorizing formulas.

It matters because matched data can fool you if you treat them like independent samples. Suppose a health class tests whether a training program changes a person's response from “does not exercise” to “exercises regularly.” If you ignore the pairing, you miss the fact that each person's before and after answers belong together. McNemar's test respects that structure and gives a cleaner answer about change.

The test also builds your interpretation skills. You are not just asking whether the percentages are different. You are asking whether the number of switches from one category to the other is balanced or lopsided. That way of thinking comes up again in repeated measures, paired observations, and any case where the same subject is measured more than once.

On problem sets and quizzes, this term helps you justify why one procedure fits and another does not. That is usually the hardest part of Honors Statistics: knowing what the data represent before you calculate anything.

How McNemar's Test connects across the course

Matched or Paired Samples

McNemar's test is built for matched or paired samples, so the pairing is not just a detail, it is the reason the test exists. If the two measurements belong to the same person or a closely matched pair, you analyze the change within each pair instead of treating the data as independent groups.

Before-After Measurements

Before-after measurements are one of the most common places McNemar's test appears. You might measure whether the same people answer a question correctly before a lesson and after a lesson, then test whether the pattern of yes/no changes is symmetric or not.

Binomial Distribution

McNemar's test uses binary outcomes, so the data often connect to binomial thinking. The test focuses on the count of discordant pairs, and the logic behind the null hypothesis comes from comparing two possible directions of change.

Null Hypothesis

The null hypothesis in McNemar's test says the two types of paired changes happen at the same rate. In other words, the number of pairs that switch from no to yes should be about the same as the number that switch from yes to no.

Normality Assumption

McNemar's test is useful because it does not rely on a normality assumption for the outcome variable. Since the response is categorical rather than numerical, the usual normal-based methods for means are not the right framework anyway.

Is McNemar's Test on the Honors Statistics exam?

A quiz or free-response problem will usually give you a paired table with yes/no data and ask which test fits. Your job is to notice that the same people, or matched subjects, were measured twice, then recognize McNemar's test as the right choice. After that, you may need to identify the null hypothesis, read the discordant pairs, and decide whether the change is statistically significant.

If the data come from a before-after study, do not jump to a two-sample proportion test. The pairing changes the whole analysis. You may also need to explain why the test is appropriate, especially if the question asks for a justification rather than just a name.

McNemar's Test vs two-proportion z-test

A two-proportion z-test compares two independent groups, while McNemar's test compares paired binary outcomes from the same subjects or matched pairs. If the data are linked pair by pair, the z-test ignores the structure and gives the wrong framework.

Key things to remember about McNemar's Test

• McNemar's test is the right choice for paired binary data, not for two independent groups.

• The test looks at whether changes go more one way than the other, like yes to no versus no to yes.

• Pairs that stay the same do not provide the main evidence for the test statistic.

• The null hypothesis says the two directions of change should happen at about the same rate.

• In Honors Statistics, this test usually appears in before-after studies, matched pairs, or case-control-style data.

Frequently asked questions about McNemar's Test

What is McNemar's Test in Honors Statistics?

McNemar's test is a hypothesis test for paired categorical data with two outcomes, such as yes/no or success/failure. You use it when the same subjects are measured twice or when subjects are matched in pairs. It checks whether the pattern of change between the two conditions is balanced or not.

When do you use McNemar's test instead of a two-proportion z-test?

Use McNemar's test when the data are paired, not independent. A two-proportion z-test is for separate groups, while McNemar's test looks at changes within each pair. If the same person is measured before and after, or if subjects are matched, McNemar's test is the better fit.

What does McNemar's test compare?

It compares the two directions of change in a binary outcome. For example, it compares how many people switch from no to yes with how many switch from yes to no. If those counts are very different, that suggests the treatment or condition may have changed the outcome.

Does McNemar's test require normality?

No. McNemar's test is used for categorical yes/no data, so normality is not the assumption you check. The important conditions are that the data are paired, the outcome is binary, and the observations within the pairs are appropriately collected.