Binomial Test — Definition, Formula & Example
A binomial test is a hypothesis test that checks whether a claimed proportion is believable. It adds up exact binomial probabilities to find the p-value, so it works even when the sample is too small for a normal approximation.
Given independent trials with a constant success probability , the binomial test evaluates using the observed number of successes as the test statistic. Under , . The p-value is the exact probability, computed from this distribution, of a result at least as extreme as in the direction of . The formula below is the right-tailed case, . For use , and a two-sided test combines both tails.
Key Formula
Where:
- = number of independent trials
- = observed number of successes
- = proportion claimed by the null hypothesis
- = number of successes, which follows a binomial distribution if the null hypothesis is true
Worked Example
Problem: Your friend claims she can taste the difference between tap water and bottled water. In 10 blind trials, she identifies the bottled water correctly 9 times. At the 0.05 significance level, is this convincing evidence that she does better than guessing?
Step 1: State the hypotheses. Guessing means . The expected number of successes is only 5, which is less than 10, so the normal approximation is not appropriate. Use the exact binomial test instead.
Step 2: Find the probability of each outcome at least as extreme as 9 correct, assuming she is guessing.
Step 3: Add these probabilities to get the p-value. Then compare it with .
Answer: The p-value is about 0.0107, which is less than 0.05, so you reject . There is convincing evidence that your friend identifies bottled water correctly more than half the time.
Why It Matters
In AP Statistics, you can use the one-proportion z-test only when both and . The binomial test covers the small-sample cases that this condition rules out. Examples include early drug trials, quality checks on a handful of parts, and taste tests.
Common Mistakes
Mistake: Using only the probability of the observed count, such as , as the p-value.
Correction: A p-value includes every outcome at least as extreme as the one you observed. For a right-tailed test, add and to get .
