The complement of an event A is the set of all outcomes in the sample space that are not in A, written A^c, with P(A^c) = 1 − P(A). On the AP Stats exam, the complement rule is the fast way to find P(at least one) and shows up in the logic of p-values, significance levels, and rejection regions.
The complement of an event A is everything in the sample space that isn't A. If A is "the employee gets selected this week," the complement A^c is "the employee does not get selected this week." Since A and A^c together cover every possible outcome and never overlap, their probabilities add to 1. That gives you the complement rule, P(A^c) = 1 − P(A), one of the most useful shortcuts in the whole course.
The reason this term earns its own page is that the complement is often easier to calculate than the event you actually care about. "What's the probability of at least one success in 10 trials?" is a nightmare to compute directly (one success, or two, or three...). But its complement, "zero successes," is a single calculation. Find P(none), subtract from 1, done. That move, flipping to the complement, is a pattern you'll reuse from probability in Unit 4 all the way into hypothesis testing in Units 6 and 7.
Complements live at the heart of probability (Unit 4), where they're defined alongside sample spaces and events. But the thinking carries straight into inference. In Topic 7.4, when you set up a one-sample t-test for a population mean (AP Stats 7.4.B), the null hypothesis H₀: μ = μ₀ and a two-sided alternative Hₐ: μ ≠ μ₀ split the possibilities into complementary claims about μ. The same logic structures significance levels (a 95% confidence level pairs with α = 0.05, and 0.95 + 0.05 = 1) and the split between the rejection region and the fail-to-reject region, which are complements of each other in the sampling distribution. If you understand complements deeply in Unit 4, the bookkeeping of inference in Units 6-8 gets much easier to follow.
Sample Space and Event (Unit 4)
A complement only makes sense relative to a sample space. A^c is literally "the sample space minus A," so listing the full sample space first is how you avoid missing outcomes when you build a complement.
Probability and the "at least one" trick (Unit 4)
P(at least one) = 1 − P(none) is the single most-tested use of complements. Anytime a problem says "at least one," your first instinct should be to compute the complement (none) and subtract from 1.
Rejection region (Units 6-7)
In a significance test, the rejection region and the fail-to-reject region are complements. They split the sampling distribution into two non-overlapping pieces, so α and 1 − α always add to 1.
Null and Alternative Hypotheses (Unit 7)
H₀: μ = μ₀ versus Hₐ: μ ≠ μ₀ is complement-style thinking applied to parameter values. The alternative covers exactly the values of μ the null leaves out, which is why a two-sided test splits its probability into both tails.
Complements show up in two main ways. First, direct probability calculations. The 2021 FRQ Q3 described a company where one of 200 employees is randomly selected each week, and the efficient path runs through the complement, P(not selected in a week) = 199/200, which you then chain across weeks to handle "at least once" questions. Second, complements appear quietly inside simulation and inference problems. If a simulation reports that 180 of 200 samples produced a symmetric distribution, you should be able to flip instantly to the 20 that didn't (a proportion of 0.10) when the question asks about the complement. On multiple choice, watch for the words "at least," "at most," and "not," since those are signals that the complement rule is the intended approach. You won't usually be asked to define complement; you'll be asked to use it without being told.
Complementary events are always mutually exclusive, but mutually exclusive events are not always complements. Disjoint just means the events can't happen together. Complements add the second requirement that the two events cover the entire sample space. Rolling a 1 and rolling a 2 are mutually exclusive, but they aren't complements because rolling a 3, 4, 5, or 6 belongs to neither. The complement of "roll a 1" is "roll anything that isn't a 1." Only for true complements can you write P(A^c) = 1 − P(A).
The complement of event A, written A^c, contains every outcome in the sample space that is not in A, and P(A) + P(A^c) = 1.
The complement rule P(A^c) = 1 − P(A) turns hard "at least one" problems into one easy calculation, since P(at least one) = 1 − P(none).
Complementary events are mutually exclusive AND exhaustive together they fill the whole sample space, which is what makes the subtraction-from-1 trick valid.
Complement thinking structures inference too. A 95% confidence level pairs with α = 0.05, and the rejection region and fail-to-reject region are complements of each other.
On the exam, the words "at least," "at most," and "not" are your cue to compute the complement instead of the event directly.
The complement of event A is the set of all outcomes in the sample space that are not in A, written A^c. Because A and A^c together cover everything, P(A^c) = 1 − P(A).
No. Complements are a special case of mutually exclusive events. Mutually exclusive only means the events can't both happen, while complements must also cover the entire sample space between them. "Roll a 1" and "roll a 2" are disjoint but not complements.
Whenever you see "at least one," "at most," or "not." P(at least one) = 1 − P(none) is almost always faster than adding up every case directly, and the 2021 FRQ Q3 (the gift card problem with 200 employees) rewarded exactly this move.
No. The definition lives in Unit 4, but the logic runs through inference. Significance level α and confidence level 1 − α are complements, and the rejection region and fail-to-reject region split the sampling distribution into complementary pieces in Units 6-7.
The complement is "2 or fewer successes" (0, 1, or 2), not "3 or fewer" and not "no successes." Getting the boundary right is the most common error, so list the outcomes on both sides and confirm they cover everything exactly once.