Two-sample t-test in AP Statistics

In AP Statistics, a two-sample t-test is a significance test that uses data from two independent random samples or two randomly assigned treatment groups to test a claim about μ₁ − μ₂, the difference between two population means (Topic 4.9).

What is Two-sample t-test?

A two-sample t-test answers one question. Is the gap between two sample means big enough that it probably isn't just random chance? You take two separate, independent groups, measure the same quantitative variable in each, and test whether the population means really differ. Per CED 4.9.A.1, this is the appropriate test whenever the parameter is a difference between two population means.

The setup has three moving parts. First, you define the parameter in context, naming the population parameters, the response variable, and both populations (4.9.A.2). For example, μ₁ − μ₂ is the true mean test score for students taught with Method A minus the true mean for Method B. Second, you write hypotheses. The null is H₀: μ₁ = μ₂ (or μ₁ − μ₂ = 0), and the alternative is one-sided (<, >) or two-sided (≠) (4.9.B.1). Third, you check conditions. You need randomization, the 10% condition for each sample (skipped for a randomized experiment), and a sample data condition showing the sampling distribution is approximately normal (4.9.C.1). The test statistic is t = ((x̄₁ − x̄₂) − 0) / √(s₁²/n₁ + s₂²/n₂). Think of it as "how many standard errors apart are the two sample means?"

Why Two-sample t-test matters in AP® Statistics

This term lives in Unit 4: Inference for Quantitative Data: Means, specifically Topic 4.9, and it supports three learning objectives. 4.9.A is about identifying the right test and parameter. 4.9.B is about writing the null and alternative hypotheses. 4.9.C is about justifying the test by verifying conditions. It's the capstone of the means unit. It combines one-sample t-procedures with the two-group comparison logic you first saw with proportions in Unit 3. It also shows up constantly because comparing two groups is the core question behind most experiments. Did the treatment group do better than the control group? That is often a two-sample t-test question.

How Two-sample t-test connects across the course

Paired t-test (Unit 4)

This is the two-sample t-test's closest neighbor and its most common mix-up. If each value in one group is naturally matched to a value in the other (same person before and after, twins, left vs. right hand), the data aren't independent. You subtract within pairs and run a one-sample t-test on the differences. The two-sample test is only for two separate, unlinked groups.

Two-sample z-test for proportions (Unit 3)

The structure is identical, with the same parameter-hypotheses-conditions-test-conclusion flow and the same H₀ of "no difference." Only the variable type changes. Categorical data means you compare proportions with z. Quantitative data means you compare means with t. If you can set up one, you can set up the other.

Combining independent random variables (Unit 2)

The standard error √(s₁²/n₁ + s₂²/n₂) comes straight from the Unit 2 rule that variances ADD when you subtract independent random variables. That's why independence between the groups matters so much. Without it, the formula breaks.

Random sampling vs. random assignment (Unit 1)

The randomization condition can be met two ways, and Unit 1 tells you what each one buys you. Random samples let you generalize to populations. Random assignment in an experiment lets you make cause-and-effect conclusions. A randomized experiment also lets you skip the 10% condition.

Is Two-sample t-test on the AP® Statistics exam?

On multiple choice, you'll mostly be asked to pick the right procedure. Some stems describe a scenario, like students randomly assigned to Teaching Method A or B with test scores recorded, and ask which test fits. Others ask which scenario would make a two-sample t-test least appropriate. The answer to that one is usually paired data or a categorical response. Other stems test whether unequal sample standard deviations are a problem (they aren't, since AP uses the unpooled test). Some give small groups, like 15 and 12 athletes, and probe the normality check or degrees of freedom.

On free response, released questions have built full inference tasks around comparing two group means. The 2025 FRQ Q6 compared reading comprehension by time of day in a study with 100 child volunteers. The 2018 FRQ Q4 involved recovery time after ACL surgery. To earn full credit, you need to do four things:

• Name the test (two-sample t-test for μ₁ − μ₂).
• Define the parameter in context, naming both populations and the response variable.
• State hypotheses using parameter symbols, not sample statistics.
• Check every condition explicitly, connecting each one to the study design. Say "random assignment" for an experiment and "random sample" for a survey.

Then finish with a conclusion in context that matches your design. Only random assignment supports a causal conclusion.

Two-sample t-test vs Paired t-test

Ask one question. Is each observation in Group 1 linked to a specific observation in Group 2? If yes, as with before/after measurements on the same subjects or matched pairs, the data are paired. You analyze the differences with a one-sample t-test on μ_diff. If the groups are separate and independent, as with two different sets of patients or two randomly assigned treatment groups, you use a two-sample t-test on μ₁ − μ₂. Running a two-sample test on paired data ignores the pairing and usually inflates the standard error, so you lose power.

Key things to remember about Two-sample t-test

• A two-sample t-test compares the means of two independent groups on a quantitative variable, and its parameter is μ₁ − μ₂.

• The null hypothesis is H₀: μ₁ = μ₂ (equivalently μ₁ − μ₂ = 0), and the alternative can be <, >, or ≠ depending on the research question.

• You must check randomization, the 10% condition for each sample (unless it's a randomized experiment), and approximate normality for each group, either through large samples or no strong skew or outliers in the data.

• AP Statistics uses the unpooled two-sample t-test, so the two groups do not need equal standard deviations.

• If the observations are naturally matched, the data are paired, and you should use a paired t-test instead of a two-sample t-test.

• Random assignment lets you conclude cause and effect, while random sampling lets you generalize to the populations.

Frequently asked questions about Two-sample t-test

What is a two-sample t-test in AP Stats?

It's a significance test that uses two independent samples or treatment groups to test a claim about μ₁ − μ₂, the difference between two population means. It's covered in Topic 4.9 of Unit 4 and tied to learning objectives 4.9.A, 4.9.B, and 4.9.C.

Do the two groups need equal standard deviations for a two-sample t-test?

No. AP Statistics uses the unpooled version, which estimates the standard error with √(s₁²/n₁ + s₂²/n₂) and handles unequal spreads fine. Sample SDs of 245 and 412 calories, for example, don't rule out the test.

How is a two-sample t-test different from a paired t-test?

A two-sample t-test compares two separate, independent groups. A paired t-test is for data where each value in one group is matched to a specific value in the other, like before/after scores on the same people. Paired data get reduced to one list of differences and tested with a one-sample t-test.

What conditions do you check for a two-sample t-test?

You check three things for each group. First, randomization, meaning random samples or random assignment. Second, the 10% condition, which is skipped for randomized experiments. Third, a normality condition, met if n ≥ 30 or if a graph of the sample data shows no strong skew or outliers.

Is the two-sample t-test on the AP Statistics exam?

Yes. It appears in multiple-choice questions asking you to identify the right procedure, and it has anchored released FRQs like 2018 Q4 (ACL surgery recovery time) and 2025 Q6 (time of day and reading comprehension). Expect to define the parameter, write hypotheses, and verify conditions in context.