The difference in two population means, μ₁−μ₂, is the true gap between the averages of two separate populations. In AP Stats you estimate it with a two-sample t-interval, (x̄₁−x̄₂) ± t*√(s₁²/n₁ + s₂²/n₂), after checking independence and normality conditions (Topic 7.6).
The difference in two population means is exactly what it sounds like. You have two populations (or two treatment groups), each with its own true mean, μ₁ and μ₂, and you want to know how far apart those means really are. You almost never know μ₁ or μ₂, so you take a sample from each group and use x̄₁−x̄₂ as your point estimate of μ₁−μ₂.
The magic that makes inference possible is the sampling distribution of x̄₁−x̄₂. If both populations are normal, or both sample sizes are bigger than 30, that sampling distribution is approximately normal with mean μ₁−μ₂ and standard deviation √(σ₁²/n₁ + σ₂²/n₂). Since you don't know the population standard deviations, you swap in s₁ and s₂ and use a t-distribution instead of z. That gives you the two-sample t-interval, (x̄₁−x̄₂) ± t*√(s₁²/n₁ + s₂²/n₂), with degrees of freedom from your calculator. The interval is a range of plausible values for the true difference, and whether 0 is inside that interval tells you whether "no difference" is plausible.
This term is the entire point of Topic 7.6 (Confidence Intervals for the Difference of Two Means) in Unit 7: Inference for Quantitative Data: Means. It runs through four learning objectives. You identify the right procedure (7.6.A), verify conditions like independent random samples, the 10% condition, and normality or n > 30 (7.6.B), compute the margin of error as t* times the standard error √(s₁²/n₁ + s₂²/n₂) (7.6.C), and build the full interval (7.6.D). It's also the payoff of the whole course design. Comparing two groups is what real studies do, whether it's a treatment group versus a control group in an experiment or two demographic groups in an observational study. One important detail from the CED: these interval formulas are not written out on the AP formula sheet in this form, so you need to know how the pieces fit together yourself.
Keep studying AP Statistics Unit 7
Confidence Interval (Unit 7)
A two-sample t-interval is just the standard confidence interval recipe, point estimate ± margin of error, applied to a difference. The point estimate becomes x̄₁−x̄₂ instead of a single x̄, and the standard error combines variability from both samples.
Hypothesis Testing (Unit 7)
The same difference shows up in two-sample t-tests, where the null hypothesis is usually μ₁−μ₂ = 0. Here's the shortcut worth remembering. If your confidence interval for μ₁−μ₂ contains 0, then "no difference" is plausible; if it doesn't, you have convincing evidence the means differ.
Sampling Distributions (Unit 5)
Everything in Topic 7.6 rests on the Unit 5 idea that statistics have their own distributions. The Central Limit Theorem is why "both n's greater than 30" rescues you when the population distributions are skewed.
Margin of Error (Units 6-7)
The margin of error here is t* times √(s₁²/n₁ + s₂²/n₂). Same logic as every other interval in the course. Bigger samples shrink it, higher confidence levels grow it, and that tradeoff is a favorite multiple-choice angle.
Multiple-choice questions love the conceptual side. A classic stem asks what happens when you drop from 95% to 90% confidence, and the answer is that fewer intervals capture the true difference in repeated sampling, but each interval gets narrower. Other MCQs test condition-checking (did they sample independently? are both n's over 30?) and interpreting whether 0 falls in the interval. On the free-response side, two-sample mean comparisons show up in full inference problems and in the investigative task. The 2025 FRQ Q6, for example, built on a psychologist's experiment comparing reading comprehension across conditions with children randomly assigned to groups. When you write the FRQ answer, you have to name the procedure (two-sample t-interval for μ₁−μ₂), check conditions explicitly, compute the interval, and interpret it in context with the word "true difference" and a direction (which group's mean is higher).
These sound identical but call for different procedures, and mixing them up is the most expensive error in Unit 7. A difference in two population means uses two independent groups, like 50 randomly assigned to morning testing and 50 to afternoon, and gets a two-sample t-interval. A mean difference uses paired data, like each child tested both morning and afternoon, where you subtract within each pair first and run a one-sample t procedure on the differences. Ask yourself one question. Are the two measurements linked person-to-person? If yes, it's paired. If the groups are separate and independent, it's two-sample.
The difference in two population means, μ₁−μ₂, is estimated by the difference in sample means, x̄₁−x̄₂.
The two-sample t-interval is (x̄₁−x̄₂) ± t*√(s₁²/n₁ + s₂²/n₂), with degrees of freedom found using technology.
Before computing, check independence (two independent random samples or a randomized experiment, plus the 10% condition when sampling without replacement) and normality (normal populations or both n₁ and n₂ greater than 30).
If 0 is inside your confidence interval, a difference of zero is plausible, so you don't have convincing evidence the population means differ.
Lowering the confidence level (say 95% to 90%) makes intervals narrower but means a smaller percentage of intervals capture the true difference in repeated sampling.
If the same subjects produce both measurements, the data are paired and you need a one-sample t procedure on the differences, not a two-sample interval.
It's the true gap μ₁−μ₂ between the averages of two separate populations or treatment groups. You estimate it with x̄₁−x̄₂ and build a two-sample t-interval around it, which is the focus of Topic 7.6.
Mostly yes. The CED notes that interval estimate formulas don't appear explicitly on the AP Statistics formula sheet, so you should know that the interval is (x̄₁−x̄₂) ± t*√(s₁²/n₁ + s₂²/n₂) and let your calculator handle the degrees of freedom.
No. It only means a difference of zero is a plausible value, so you lack convincing evidence of a difference. A confidence interval can never prove the means are exactly equal.
Two means requires two independent groups, like separate random samples or random assignment to two treatments. Paired data links measurements, like before-and-after scores for the same person, and uses a one-sample t procedure on the differences instead.
Two things. Independence, meaning independent random samples or a randomized experiment plus n₁ ≤ 10% of N₁ and n₂ ≤ 10% of N₂ when sampling without replacement. And approximate normality, meaning normal populations or both sample sizes greater than 30.