A confidence interval is a range of plausible values for an unknown population parameter, built as point estimate ± margin of error from sample data. In repeated random sampling, approximately C% of intervals constructed this way will capture the true parameter (proportion, mean, or slope).
A confidence interval answers the question "based on my sample, what values of the population parameter are plausible?" You can't know the true population proportion, mean, or regression slope exactly, so instead of one guess you give a range. Every confidence interval on the AP exam has the same skeleton: point estimate ± (critical value)(standard error). The point estimate is your sample statistic (p̂, x̄, x̄₁ − x̄₂, or b), and the critical value times the standard error is the margin of error.
The confidence level (usually 90%, 95%, or 99%) describes the method, not any single interval. If you took repeated random samples of the same size and built an interval from each one, approximately C% of those intervals would capture the true parameter. Any one interval either contains the parameter or it doesn't. That's why the AP-approved interpretation says "we are 95% confident that the interval captures the true population mean," never "there is a 95% probability the mean is in this interval." Before you calculate anything, you have to verify conditions for independence (random sampling, the 10% condition when sampling without replacement) and approximate normality of the sampling distribution.
Confidence intervals are one of the two pillars of statistical inference, and they run through three full units of the course. Unit 6 covers z-intervals for one proportion (LOs 6.2.A-E, 6.3.A-C) and for a difference of two proportions (6.8.A-D, 6.9.A-B). Unit 7 covers t-intervals for one mean, including matched pairs (7.2.A-E, 7.3.A-C), and for a difference of two means (7.6.A-D, 7.7.A-C). Unit 9 extends the same logic to the slope of a regression model (9.2.A-D, 9.3.A-C). Unit 5 sets up the foundation, since a sample statistic is just a point estimator with variability you can model (5.4.A-B). The procedures change, but the structure, the conditions checklist, the interpretation template, and the width relationships (bigger n means narrower interval; higher confidence level means wider interval) are the same every time. Learn the pattern once and you've learned a third of the course.
Keep studying AP Statistics Unit 7
Margin of Error (Units 6-7)
The margin of error is half the confidence interval. It equals the critical value times the standard error, and it tells you how far your sample statistic is likely to sit from the true parameter. The width of the interval is exactly twice the margin of error, a fact MCQs love to test.
Point Estimate and Sampling Distributions (Unit 5)
A confidence interval is built around a point estimate, and its margin of error comes straight from the sampling distribution of that statistic. Unit 5 is where you learn why p̂ and x̄ are unbiased estimators with predictable variability. A confidence interval is basically a sampling distribution turned into a usable range.
Significance Tests (Units 6-7)
Intervals and tests are two views of the same inference. If a 95% confidence interval for a difference of proportions doesn't contain 0, that's evidence the two proportions actually differ, matching a two-sided test at α = 0.05. FRQs often let you justify a claim either way, but you have to say what the interval contains (or doesn't) in context.
Confidence Interval for a Regression Slope (Unit 9)
The same formula structure, b ± t*(SE_b), estimates the true slope β of the population regression line. The conditions change (linearity, equal standard deviation of y across x), but the interpretation template and the sample-size effect on width carry over directly from Units 6 and 7.
t-Distributions (Unit 7)
When you don't know σ (which is basically always for means), you swap in s and use t* instead of z*. The t-distribution has fatter tails than the normal curve, which makes the interval slightly wider, an honest price for estimating the standard deviation from the sample.
Confidence intervals show up everywhere. MCQs test the width relationships (what happens when n quadruples, or when confidence rises from 90% to 99%), the correct interpretation of "95% confident," condition checks, and reading intervals off computer output. FRQs typically ask you to do the full four-step procedure, which means naming the interval (one-sample z-interval for a proportion, two-sample t-interval for a difference of means, etc.), verifying conditions, calculating, and interpreting in context. The 2018 FRQ had you build and interpret an interval for the proportion of students who recycle, and the 2019 apartment-rent FRQ used an interval for a mean. Graders want the population, the parameter, and the context named explicitly, so "we are 95% confident the interval from 0.58 to 0.70 captures the true proportion of all students at this school who regularly recycle" earns credit while "the answer is between 0.58 and 0.70" does not. The interval formulas aren't on the formula sheet as such, but you can build every one from the point estimate ± (critical value)(SE) pattern using the standard errors that are provided.
The confidence interval is the actual range you compute from one sample, like (0.42, 0.54). The confidence level (95%, say) describes the long-run success rate of the method that produced it. The single most common AP error is saying "there's a 95% probability the parameter is in my interval." Once the interval is built, the parameter is either in it or not. The 95% refers to repeated sampling, meaning about 95% of all intervals built this way would capture the true parameter.
Every AP confidence interval has the form point estimate ± (critical value)(standard error), whether you're estimating a proportion, a mean, a difference, or a regression slope.
The confidence level describes the method, not one interval. In repeated random sampling, approximately C% of intervals constructed this way capture the true parameter.
A correct interpretation references the sample taken, names the parameter, and describes the population in context. Skipping context costs points on FRQs.
Increasing sample size makes the interval narrower (width is proportional to 1/√n for a single proportion or mean), while increasing the confidence level makes it wider.
Always verify conditions before calculating, including randomness, the 10% condition when sampling without replacement, and approximate normality of the sampling distribution.
You can justify a claim with an interval by checking whether it contains a key value. An interval for a difference that excludes 0 supports the claim that a real difference exists.
It's a range of plausible values for an unknown population parameter, calculated as point estimate ± margin of error. For example, p̂ ± z*√(p̂(1−p̂)/n) estimates a population proportion, and x̄ ± t*(s/√n) estimates a population mean.
No, and this misconception loses points on every exam. Once the interval is calculated, the parameter is either in it or it isn't. The 95% means that in repeated random sampling, about 95% of intervals built with this method would capture the true parameter.
An interval estimates the plausible values of a parameter; a test checks whether a specific hypothesized value is plausible. They're linked, since a 95% two-sided interval that excludes the null value (like 0 for a difference of proportions) corresponds to rejecting H₀ at α = 0.05.
Use z* for proportions because the standard error comes from p̂ itself. Use t* for means and regression slopes because you're estimating σ with the sample standard deviation s, which adds extra uncertainty that the t-distribution's fatter tails account for.
Larger samples give narrower intervals because the standard error shrinks. For a single mean or proportion, width is proportional to 1/√n, so quadrupling the sample size cuts the width in half. This relationship is a favorite MCQ setup.