This skill brings together inference for means and proportions so you can pick the right procedure for a problem. The main moves are deciding between means () and proportions (), figuring out how many samples you have, checking conditions, running the math, and writing a clear conclusion in context.
Why This Matters for the AP Statistics Exam
By this point you have a full toolbox: one-sample and two-sample t-procedures for means, z-procedures for proportions, and matched pairs methods. The exam expects you to choose correctly when no one tells you which test to use. Multiple-choice questions often test whether you can match a scenario to the right procedure, interpret a p-value, or draw a conclusion from one. Free-response questions frequently ask "do the data give convincing evidence" (a significance test) or ask you to construct and interpret a confidence interval. Picking the wrong procedure or misreading the setup is one of the easiest ways to lose points, so this is a high-value skill to practice.
Key Takeaways
- Ask two questions first: means (t) or proportions (z), and one sample or two? Those answers point you to the correct procedure.
- Matched pairs data uses a one-sample t-procedure on the differences, not a two-sample test.
- A p-value is the probability of getting a result as extreme as your sample, assuming the null hypothesis is true.
- Compare the p-value to alpha to decide: reject H0 when p is less than or equal to alpha, fail to reject when p is greater than alpha.
- Never "accept" a hypothesis and never state hypotheses using sample statistics; use population parameters.
- Always check conditions and write your conclusion in context.
How to Use This on the AP Statistics Exam
MCQ
Multiple-choice inference questions usually test one of three things: selecting the correct procedure, interpreting a p-value, or drawing a conclusion from a p-value.
Selecting the correct procedure
Ask yourself two questions:
- Am I dealing with means (t-procedures) or proportions (z-procedures)?
- Do I have one sample or two?
These two answers tell you whether to run a one-sample or two-sample, mean or proportion, test or interval.
A matched pairs t-test (also called a dependent samples t-test) compares the means of two related groups, where each subject in one group is paired with a subject in the other. This often shows up in experiments where every unit receives both treatments (for example, a drug and a placebo) and you measure the difference for each unit. You analyze the differences as a single sample.
A two-sample t-test compares the means of two independent groups, where subjects in one group are not related to subjects in the other (for example, men and women). Do not confuse matched pairs with two independent samples.
If you need to compare more than two proportions or look at categorical data in a table, a chi-square test may be the right tool. A chi-square test compares observed frequencies with expected frequencies in a two-way table and is used for questions about categorical data. You will work with chi-square procedures in the next unit, so just recognize when a problem points that direction.
Interpreting a p-value
A p-value is the probability of obtaining a sample result as extreme as yours, assuming the null hypothesis is true (that the true mean or proportion equals the value stated in H0).
Example: In a test where H0: p = 0.2 and Ha: p < 0.2, you collect a sample of 100 with p-hat = 0.15, and the test gives a p-value of 0.11. Interpret it.
A p-value of 0.11 means that if the true proportion really were 0.2, you would get a sample of 100 with a success rate of 0.15 or lower about 11% of the time.
Drawing conclusions
To conclude, compare the p-value to the significance level alpha:
| Decision | Meaning |
|---|---|
| p ≤ alpha: Reject H0 | We have convincing evidence of Ha (in context). |
| p > alpha: Fail to reject H0 | We do not have convincing evidence of Ha (in context). |
Remember: you never "accept" H0 or Ha.
Free Response
Free-response inference questions usually come in one of two forms:
- "Do the data give convincing evidence..." (significance test)
- "Construct and interpret a ___% confidence interval" (confidence interval)
Both follow the same four-step structure. Showing each step clearly is important for full credit.
(1) State the parameters and hypotheses
- For a confidence interval, state the parameter(s) you are estimating for the population(s).
- For a significance test, write your hypotheses and define your parameters. H0 always uses an equal sign, and Ha always uses an inequality (<, >, or not equal to). Write hypotheses in terms of population parameters, not sample statistics.
(2) Plan the problem
Check your conditions for inference: random, independent, and approximately normal. The conditions are similar for intervals and tests, but the details depend on whether your data are categorical or quantitative. For means, the normal condition is met if the population is normal or the sample size is large enough (often n > 30); with smaller samples, check that the data are free from strong skewness and outliers. When sampling without replacement, also check that the sample is at most 10% of the population.
(3) Do the math
- Identify the specific test or interval you are running (this is usually the function you select on your calculator). Write it down.
- Then report your results:
- Confidence intervals: the interval is enough.
- Significance tests: the test statistic, p-value, and degrees of freedom (when needed).
(4) Conclusion
Use these templates, filled in with context:
- For confidence intervals: "I am _% confident that the true _____ of __________ is between (, __)."
- For significance tests: "Since the p-value (< or >) alpha, I (reject / fail to reject) H0. There (is / is not) convincing evidence of Ha (in context)."
Common Misconceptions
- Mixing up matched pairs and two-sample t-tests. If each subject is paired with another or each unit gets both treatments, analyze the differences as one sample. Independent groups call for a two-sample t-test.
- Using z when you should use t. For means with an unknown population standard deviation, use t-procedures. A z-test for a mean is only appropriate when the population standard deviation is known.
- Writing hypotheses with sample statistics. Hypotheses are always about population parameters (mu, p), never about x-bar or p-hat.
- Saying you "accept" the null. You either reject H0 or fail to reject it. Failing to reject is not proof that H0 is true.
- Misreading the p-value. It is the probability of a result as extreme as yours assuming H0 is true, not the probability that H0 is true.
- Skipping conditions. Even if your calculator gives an answer, you still need to check random, independent, and normal, or you lose the reasoning that supports your conclusion.
- Vague conclusions. Always tie your decision back to the context of the problem instead of just saying "reject" or "fail to reject."
Related AP Statistics Guides
- Unit 7 Overview: Means
- 7.2 Constructing a Confidence Interval for a Population Mean
- 7.1 Introducing Statistics: Should I Worry About Error?
- 7.4 Setting Up a Test for a Population Mean
- 7.3 Justifying a Claim About a Population Mean Based on a Confidence Interval
- 7.6 Confidence Intervals for the Difference of Two Means
Frequently Asked Questions
How do I choose the right inference procedure in AP Statistics?
First decide whether the parameter is a mean or a proportion. Then decide whether the data come from one sample, two independent samples, or matched pairs. Those choices point you to the correct t, z, or matched-pairs procedure.
When do you use a t-procedure instead of a z-procedure?
Use t-procedures for inference about means when the population standard deviation is unknown. Use z-procedures for inference about proportions, where the normal approximation is based on counts of successes and failures.
What is the difference between matched pairs and two independent samples?
Matched pairs data have linked observations, such as before-and-after measurements on the same subjects. Analyze the differences with a one-sample t-procedure. Two independent samples compare separate groups with no pairing.
What conditions should I check for AP Stats inference?
Check random selection or assignment, independence, and an appropriate normal or large-sample condition. The exact normal condition depends on whether you are working with means, proportions, or differences.
How do you write a conclusion for a significance test?
Compare the p-value to alpha. If p is less than or equal to alpha, reject H0 and say there is convincing evidence for Ha in context. If p is greater than alpha, fail to reject H0 and say there is not convincing evidence.
How do you write a confidence interval conclusion?
State that you are the given percent confident that the true population parameter lies between the interval endpoints, and include the context of the problem. Name the parameter, not just the numbers.