The difference of two population means, written μ1 − μ2, is the parameter that tells you how far apart the true averages of two groups are; in AP Stats Topic 7.7 you estimate it with a two-sample t confidence interval and check whether the interval supports a claim (especially whether it contains 0).
The difference of two population means is exactly what it sounds like. You have two populations (two fire stations, two teaching methods, two species of fish), each with its own true mean. Subtract one mean from the other and you get the parameter μ1 − μ2. You almost never know this number, so you estimate it using the difference of your two sample means and build a confidence interval around that estimate.
The order of subtraction matters. If you define the difference as (northern − southern), then a positive interval means the northern mean is bigger, a negative interval means it's smaller, and an interval containing 0 means you can't rule out 'no difference at all.' Per UNC-4.Z.2, your interpretation has to mention the samples you took and the populations they represent, with the direction of subtraction stated clearly. 'We are 95% confident the interval captures the true difference in mean response times (northern − southern)' is the kind of sentence the exam wants.
This parameter is the centerpiece of Topic 7.7 in Unit 7 (Inference for Quantitative Data: Means). Three learning objectives hang on it. AP Stats 7.7.A asks you to interpret a confidence interval for a difference of population means, including the repeated-sampling idea in UNC-4.Z.1 (about C% of intervals built this way would capture the true difference). AP Stats 7.7.B asks you to justify a claim using that interval, which usually comes down to one question: is 0 inside the interval? Per UNC-4.AA.1, the interval itself is your evidence. AP Stats 7.7.C covers the sample-size effect, where bigger samples shrink the interval's width when everything else stays the same (UNC-4.AB.1). Two-group comparisons are also how real experiments get analyzed, so this concept connects directly back to study design from Unit 3.
Keep studying AP Statistics Unit 7
Confidence Interval (Unit 7)
The difference of two population means is the parameter; the confidence interval is the tool you use to estimate it. Same logic as a one-sample interval, just with two samples feeding the point estimate and the standard error.
Null Hypothesis (Units 6-7)
The typical null hypothesis for two means is μ1 − μ2 = 0, meaning no difference. A confidence interval and a two-sided significance test agree here. If 0 falls outside your 95% interval, a test at α = 0.05 would reject that null.
Standard Error (Units 5-7)
Two samples means two sources of sampling variability, so the standard error for x̄1 − x̄2 combines the variability from both groups. That's why two-sample intervals tend to be wider than one-sample intervals built from similar data.
Margin of Error (Units 6-7)
The interval's width is twice the margin of error, and UNC-4.AB.1 says that width shrinks as sample sizes grow. Quadrupling both sample sizes roughly cuts the margin of error in half, the same square-root relationship you saw with proportions in Unit 6.
No released FRQ uses the phrase 'difference of two population means' verbatim, but comparing two means is a classic inference setup on both sections of the exam. Multiple-choice questions hand you an interval like (−2.3, 5.1) and ask what it means that 0 is inside, or ask how doubling the sample sizes changes the width. Free-response inference questions expect the full routine in context. Define the parameter with the direction of subtraction stated (μN − μS, the true difference in mean response times), check conditions, build the two-sample t-interval, and interpret it. The most common point lost is a generic interpretation that never names the populations or the variable. UNC-4.Z.2 makes context non-negotiable. Also watch for the claim-justification twist from 7.7.B, where the question asks 'is there convincing evidence the means differ?' and your answer must point to whether the interval contains 0.
μ1 − μ2 is the parameter, a fixed unknown number about the populations. The difference in sample means, x̄1 − x̄2, is the statistic you actually calculate from data, and it changes from sample to sample. The statistic is the center of your confidence interval; the parameter is the thing the interval is trying to capture. Saying 'there's a 95% chance μ1 − μ2 is in my interval' confuses the two and costs you interpretation credit.
The difference of two population means, μ1 − μ2, is a fixed unknown parameter that you estimate with the difference of sample means and a two-sample t confidence interval.
If the confidence interval contains 0, you do not have convincing evidence that the two population means differ; if the entire interval is positive or negative, you do.
Always state the direction of subtraction, because the interval (1.2, 4.6) for (A − B) means population A's mean is plausibly 1.2 to 4.6 units higher than B's.
A correct interpretation references the samples taken and the populations they represent, not just 'the true difference.'
When everything else stays the same, increasing the sample sizes makes the interval for the difference of two means narrower.
The C% confidence level describes the method, meaning about C% of intervals built from repeated random samples of the same sizes would capture the true difference.
It's the parameter μ1 − μ2, the gap between the true averages of two populations. In Topic 7.7 you estimate it with a two-sample t confidence interval and use that interval to judge claims about whether and how much the means differ.
No. An interval containing 0 means you lack convincing evidence of a difference, not that the difference is exactly 0. The true difference could still be any value inside the interval, including small nonzero ones.
Two population means come from two independent groups, like two separate fire stations. A paired mean difference comes from matched or repeated measurements on the same subjects, like before-and-after scores, which you analyze as one sample of differences. Using the two-sample procedure on paired data is a classic AP error.
Larger samples reduce the standard error of x̄1 − x̄2, which shrinks the margin of error. Per UNC-4.AB.1, when everything else stays the same, increasing the sample sizes decreases the interval's width.
No. The parameter is a fixed number, so any single interval either captures it or it doesn't. The 95% refers to the method: in repeated random sampling, about 95% of intervals built this way would capture the true difference (UNC-4.Z.1).