AP Stats Hypothesis Testing: Explained Simply
AP Stats hypothesis testing explained: hypotheses, conditions, p-values, Type I and II errors, power, and a worked example that earns full FRQ credit.
Hypothesis testing in AP Statistics asks one question: is the evidence in your sample strong enough to reject a claim about a population? You assume the null hypothesis is true, measure how unusual your sample result would be under that assumption (the p-value), and reject the null only if the p-value falls below your significance level. Master the four-step structure and the precise wording, and significance tests become some of the most predictable points on the exam.
What are the null and alternative hypotheses?
- The null hypothesis (H₀) is the claim of "no effect" or "no difference." It always uses an equals sign and a specific parameter value.
- The alternative hypothesis (Hₐ) is what you are looking for evidence of. It uses <, >, or ≠.
Hypotheses are always about population parameters (p or μ), never sample statistics (p̂ or x̄).
Example. A school says 70% of its AP students score a 3 or higher. A parent suspects the true rate is lower.
- H₀: p = 0.70
- Hₐ: p < 0.70
where p is the true proportion of the school's AP exam takers who score 3 or higher. Defining the parameter in words is part of full credit.
One-sided or two-sided?
Use a one-sided alternative (< or >) only when the question specifies a direction before seeing the data — "lower," "higher," "improved." Use ≠ when the question asks whether something is "different" or "has changed."
What are the four steps of a significance test?
- State the hypotheses, define the parameter, and give α.
- Plan: name the test and check conditions.
- Do: compute the test statistic and p-value.
- Conclude: compare the p-value to α and state your conclusion in context.
The conditions (Random, 10%, and Normal/Large Counts) are the same ones you check for confidence intervals, with one twist for proportions: in a test, you check Large Counts using the hypothesized value p₀, not p̂.
Worked example: one-sample z-test for a proportion
A random sample of 150 of the school's AP exam takers shows 96 scored 3 or higher. Test the parent's claim at α = 0.05.
State. H₀: p = 0.70; Hₐ: p < 0.70, where p is the true proportion of the school's AP exam takers scoring 3 or higher. α = 0.05.
Plan. One-sample z-test for p.
- Random: random sample.
- 10%: 150 is less than 10% of all the school's AP exam takers over the period studied (assume this is reasonable).
- Large Counts: 150(0.70) = 105 ≥ 10 and 150(0.30) = 45 ≥ 10.
Do. p̂ = 96/150 = 0.64.
Standard deviation of p̂ under H₀: √(0.70 × 0.30 / 150) = √0.0014 ≈ 0.0374
z = (0.64 − 0.70) / 0.0374 ≈ −1.60
p-value = P(Z ≤ −1.60) ≈ 0.055
Conclude. Because the p-value of 0.055 is greater than α = 0.05, we fail to reject H₀. We do not have convincing evidence that the true proportion of the school's AP exam takers scoring 3 or higher is less than 0.70.
Notice the conclusion: borderline evidence is still not convincing at the chosen level. You never "accept" the null.
What does a p-value actually mean?
The p-value is the probability of getting a sample result at least as extreme as the one observed, assuming the null hypothesis is true.
Template for interpreting it in context:
Assuming the true proportion of the school's AP exam takers who score 3 or higher is 0.70, there is about a 0.055 probability of getting a sample proportion of 0.64 or lower in a random sample of 150 students by chance alone.
What a p-value is not
- It is not the probability that the null hypothesis is true.
- It is not the probability that the result happened by chance.
- A large p-value does not prove the null is true; it means the data are consistent with it.
How do you write the conclusion?
Use this two-part rule:
| p-value vs. α | Decision | Conclusion wording |
|---|---|---|
| p-value < α | Reject H₀ | We have convincing evidence that [Hₐ in context]. |
| p-value ≥ α | Fail to reject H₀ | We do not have convincing evidence that [Hₐ in context]. |
Always link your decision to the p-value explicitly ("because 0.012 < 0.05...") and always state the conclusion in terms of the alternative hypothesis, in context.
What are Type I and Type II errors?
Because you decide based on a sample, you can be wrong in two ways.
| H₀ actually true | H₀ actually false | |
|---|---|---|
| Reject H₀ | Type I error | Correct decision |
| Fail to reject H₀ | Correct decision | Type II error |
- Type I error: rejecting a true null — finding an effect that is not there. Its probability equals α.
- Type II error: failing to reject a false null — missing a real effect.
On the exam, you will often be asked to describe an error in context and give a consequence. For the example above:
- Type I: concluding the school's rate is below 70% when it really is 70%. Consequence: the school might be unfairly criticized or change a program that works.
- Type II: failing to find evidence the rate is below 70% when it really is lower. Consequence: struggling students might not get extra support.
What is power, and how do you increase it?
Power is the probability of correctly rejecting H₀ when a specific alternative is true. Power = 1 − P(Type II error).
| To increase power | Why it works |
|---|---|
| Increase sample size | Reduces variability of the statistic |
| Increase α | Makes it easier to reject H₀ (but raises Type I risk) |
| True parameter farther from the null value | A bigger effect is easier to detect |
| Reduce variability (better design, blocking) | Sharper estimates |
Increasing sample size is the answer graders usually want, because it raises power without increasing the Type I error rate.
Which test should you use?
Choosing the right procedure is half the battle. Ask: proportions or means? One sample, two independent samples, or paired data?
| Situation | Test |
|---|---|
| One proportion | One-sample z-test for p |
| Two independent proportions | Two-sample z-test for p₁ − p₂ |
| One mean | One-sample t-test for μ |
| Paired data (before/after, matched pairs) | Paired t-test for μ_diff |
| Two independent means | Two-sample t-test for μ₁ − μ₂ |
| Categorical distribution or association | Chi-square tests |
| Slope of a regression line | t-test for slope |
Paired versus two-sample is a classic trap: if each value in one group is naturally linked to a value in the other (the same student before and after), use paired procedures. Our AP Statistics inference cheat sheet has every test on one page.
How Ivystar can help
Significance tests reward exact wording, and small phrasing slips cost real points. Our tutors run students through dozens of graded inference FRQs until the structure is automatic — part of why 91% of Ivystar AP students score a 5. Explore our AP Exam Prep program, estimate your score with the AP Statistics score calculator, and book a free 20-minute strategy call.
Frequently asked questions
Can you ever accept the null hypothesis in AP Stats?+
No. If the p-value is not below alpha, you fail to reject the null, meaning the data do not provide convincing evidence for the alternative. That is not proof the null is true.
Why do you use p0 instead of p-hat when checking conditions in a test?+
A significance test is carried out assuming the null hypothesis is true, so the expected counts and standard deviation are computed using the hypothesized value p0.
What significance level should I use if none is given?+
Use 0.05, the conventional default, and state it explicitly in your answer.
How is power related to Type II error?+
Power equals 1 minus the probability of a Type II error. A test with higher power is less likely to miss a real effect.
What is the most common mistake on hypothesis testing FRQs?+
Writing conclusions without context or without linking the decision to the p-value. Always compare the p-value to alpha and state the conclusion about the alternative hypothesis in the words of the problem.
Get a plan built for your student
Book a free 20-minute strategy call with Ivystar's founder. Get an A — or we work for free.