A 95% confidence interval is a range of plausible values for a population parameter, built as point estimate ± margin of error, where the method captures the true parameter in about 95% of all possible random samples of the same size.
A 95% confidence interval is your best answer to the question "what is the true population value, given that I only have a sample?" You take your point estimate (like a sample mean x̄ or a sample proportion p̂), then add and subtract a margin of error to create a range of plausible values for the parameter. For a population mean, that's x̄ ± t*(s/√n), the one-sample t-interval from Topic 7.2.
The "95%" describes the method, not any single interval. If you took many random samples of the same size and built an interval from each one, about 95% of those intervals would capture the true parameter. Your one interval either contains the truth or it doesn't. You just don't know which, so you say you're "95% confident" in the process that produced it. That's the interpretation the CED requires (AP Stats 6.9.A), and graders look for the repeated-sampling idea plus context about the population.
The 95% confidence interval is the workhorse of inference, showing up in Unit 6 (proportions) and Unit 7 (means). In Topic 6.9, you use an interval for a difference of proportions to justify a claim (AP Stats 6.9.A and 6.9.B). The big move is checking whether 0 falls inside the interval. If it does, you can't rule out "no difference." In Topic 7.2, you build the interval itself for a mean, which means describing t-distributions (7.2.A), picking the right procedure (7.2.B), verifying random sampling, the 10% condition, and normality (7.2.C), computing margin of error t*(s/√n) (7.2.D), and calculating the full interval (7.2.E). Inference questions like these anchor the back half of the course and are nearly guaranteed FRQ territory.
Margin of Error (Units 6-7)
The margin of error is the entire "±" part of the interval. For a 95% t-interval it equals t* times s/√n, so a bigger sample shrinks the interval and a higher confidence level widens it. The interval is just the point estimate with this buffer attached on both sides.
Confidence Level (Units 6-7)
95% is one choice of confidence level, not the only one. Bump it to 99% and t* gets bigger, so the interval widens. Drop to 90% and it narrows. Confidence and precision trade off against each other, which is a favorite MCQ idea.
Critical Value (Unit 7)
The critical value t* sets how many standard errors you reach out from x̄. It comes from a t-distribution with n-1 degrees of freedom because you're using s instead of σ. The t-curve has fatter tails than the normal curve, and as degrees of freedom grow, it looks more and more normal (AP Stats 7.2.A).
Independence (Units 4-7)
Before you build any interval, you have to verify conditions. Independence comes from random sampling (or random assignment) plus the 10% condition when sampling without replacement. Skipping the condition check on an FRQ costs you, even if your math is perfect.
Confidence intervals get tested two ways. First, construction questions ask you to name the procedure ("one-sample t-interval for a mean"), check conditions, compute the interval, and interpret it in context. The 2022 FRQ gave a random sample of 920 teenagers and asked about a proportion, classic interval setup. Second, claim-justification questions hand you a finished interval and ask what conclusion is supported. Multiple-choice stems frequently give an interval for a difference of proportions, like (-0.07, 0.01), and ask whether a claim such as "Candidate A has more support" is justified. The key check is whether 0 is in the interval. If it is, the data don't give convincing evidence of a difference. Watch for trap answers that claim causation from observational data, even when the interval excludes 0. Also note the formula sheet doesn't list interval formulas directly, but you can build them from the test statistic and standard error formulas it does provide.
Once you've computed a specific interval like (0.55, 0.63), the true parameter is either in it or not. There's no 95% probability about it. The 95% refers to the long-run capture rate of the method. Saying "there's a 95% chance μ is between 0.55 and 0.63" is the single most common interpretation error and loses credit on FRQs. Say instead that you're 95% confident the interval captures the true parameter, because the method works about 95% of the time.
A 95% confidence interval is point estimate ± margin of error, and for a population mean that means x̄ ± t*(s/√n).
The correct interpretation is about the method, not one interval. In repeated random sampling, about 95% of intervals built this way would capture the true parameter.
To justify a claim about a difference of proportions or means, check whether 0 is inside the interval. If 0 is included, you cannot conclude there's a difference.
Always verify conditions first. You need random sampling, n ≤ 10% of the population when sampling without replacement, and an approximately normal sampling distribution (n > 30 or a roughly symmetric sample).
Use t* instead of z* for means because σ is unknown and you're estimating it with s, which puts more area in the tails of the distribution.
Higher confidence means a wider interval, and a larger sample size means a narrower one. You can't improve both confidence and precision for free.
It's a range of plausible values for a population parameter, calculated as point estimate ± margin of error from sample data. The 95% means that if you repeated the sampling process many times, about 95% of the intervals you'd build would capture the true parameter.
No, and this is the misconception AP graders watch for. A specific interval either contains the parameter or it doesn't. The 95% describes the success rate of the method over many samples, so you say you're "95% confident," not that there's a 95% probability.
The confidence interval is the actual range of values, like (0.42, 0.58). The confidence level is the 95% (or 90%, or 99%) that tells you how reliable the method is in the long run. Raising the level widens the interval because t* or z* gets larger.
Check whether the claimed value falls inside the interval. For a difference of proportions, if 0 is inside the interval, the data don't provide convincing evidence of a difference (AP Stats 6.9.B). If the entire interval is above or below 0, you have evidence for a difference in that direction.
Use t* with n-1 degrees of freedom whenever you're estimating a population mean and σ is unknown, which on the AP exam is basically always for means. Proportions use z* because the standard error formula doesn't require estimating a separate standard deviation.