Setting Up a Test for a Population Mean - AP Statistics

SECTION 1

Historical Context & Motivation

Statistical hypothesis testing grew out of a practical need: how can a researcher distinguish a genuine effect from mere sampling variability? Before formal tests existed, scientists relied on informal judgment when deciding whether experimental data supported a claim about a population parameter. The development of significance testing gave researchers a structured, reproducible framework for making such decisions—one rooted in probability theory rather than intuition.

1908 • Student's t-Distribution

William Sealy Gosset, publishing under the pseudonym "Student," derived the t-distribution while working at the Guinness Brewery. His work addressed the problem of making inferences from small samples when the population standard deviation is unknown.

1925 • Fisher's Framework

Ronald Fisher formalized the concept of a null hypothesis and the p-value in his landmark book Statistical Methods for Research Workers, establishing the logic of testing a claim by assuming it true and measuring how surprising the data are.

1933 • Neyman–Pearson Theory

Jerzy Neyman and Egon Pearson introduced the alternative hypothesis and the concepts of Type I and Type II errors, transforming hypothesis testing into a decision-theoretic framework with explicit error-rate control.

Modern Era • AP Statistics Curriculum

The modern AP Statistics course synthesizes Fisher's p-value approach with the Neyman–Pearson decision framework, requiring students to state hypotheses, check conditions, compute test statistics, and draw conclusions in context.

The central question this lesson addresses is deceptively simple: given a claim about a population mean μ, how do we translate that claim into a pair of competing hypotheses that can be evaluated with sample data? This "setup" phase—writing hypotheses, identifying the parameter, choosing the correct test, and verifying conditions—is the foundation upon which every subsequent calculation and conclusion rests. An error here propagates through the entire analysis, so mastering the setup is essential before turning to test statistics and p-values.

SECTION 2

Core Principles & Definitions

Setting up a hypothesis test for a population mean requires you to clearly define the parameter under investigation, articulate two competing claims, and verify that the sampling conditions justify the use of a t-procedure. Each of these components must be stated precisely and in context—generic or vague language is one of the most common reasons students lose points on the AP exam. The following grid introduces the four foundational ideas you need before writing a single formula.

Parameter of Interest

The population mean μ must be defined in context. For example, "μ = the true mean body temperature (in °F) of all healthy adults." Always use μ, never x̄, when stating what you are testing.

Null Hypothesis (H₀)

The null hypothesis is a statement of no effect or no difference. It always takes the form H₀: μ = μ₀, where μ₀ is a specific claimed value. This serves as the assumption to be tested.

Alternative Hypothesis (Hₐ)

The alternative hypothesis represents the research claim or suspicion. It takes one of three forms: μ < μ₀ (left-tailed), μ > μ₀ (right-tailed), or μ ≠ μ₀ (two-tailed). The direction is chosen before seeing the data.

Conditions for Inference

Three conditions must be verified: (1) Random—data come from a random sample or randomized experiment; (2) Independence—the 10% condition, n ≤ 0.10N; (3) Normal/Large Sample—the population is approximately normal or n ≥ 30 (CLT).

✦

KEY TAKEAWAY

Think of the null hypothesis as the "status quo" position in a courtroom trial—it is presumed true until the evidence (data) convinces us otherwise. Just as a prosecutor must specify the exact charge before a trial begins, you must specify H₀ and Hₐ before collecting or analyzing data. The alternative hypothesis represents the claim you hope to find evidence for, while the null represents the default position of no change or no difference. Getting the direction of Hₐ right is analogous to choosing whether you are testing for guilt on a specific charge—it shapes every subsequent step of the analysis.

SECTION 3

Visual Explanation: Anatomy of the Test Setup

The following diagram illustrates the complete process of setting up a one-sample t-test for a population mean. It traces the flow from an initial research question through defining the parameter, writing hypotheses, selecting the tail direction, and verifying conditions—all before any computation takes place.

The flowchart begins with a research question and proceeds through five stages: defining the parameter μ in context, writing H₀ and Hₐ, selecting the tail direction based on the research question (not the data), and verifying the three conditions for a one-sample t-test. Notice that the tail direction must be determined by the research question before examining sample results.

Several points in the diagram deserve special emphasis. First, the parameter definition box explicitly requires you to state what μ represents in the context of the problem and to include units—this is not optional on the AP exam. Second, notice that the three branch options for Hₐ are mutually exclusive; the direction is dictated by the research question, not by preliminary inspection of x̄. Third, the conditions check at the bottom must be addressed with reference to the specific problem scenario—merely listing the names of the conditions without justification earns minimal credit on free-response questions.

SECTION 4

Mathematical Framework

Although the focus of this lesson is on the setup rather than computation, understanding the mathematical structure of the one-sample t-test clarifies why each element of the setup matters. The test statistic, the degrees of freedom, and the conditions all flow directly from the assumptions encoded in your hypotheses.

NULL AND ALTERNATIVE HYPOTHESES

H₀: μ = μ₀ Hₐ: μ < μ₀ or μ > μ₀ or μ ≠ μ₀

μ is the true population mean; μ₀ is the hypothesized value (a specific number drawn from the problem context). The null always uses "=" while the alternative uses a strict inequality.

ONE-SAMPLE t TEST STATISTIC

t = (x̄ − μ₀) / (s / √n)

x̄ = sample mean, s = sample standard deviation, n = sample size, μ₀ = hypothesized population mean. The denominator s / √n is the standard error of x̄. This statistic follows a t-distribution with df = n − 1 when H₀ is true and conditions are met.

DEGREES OF FREEDOM

df = n − 1

The degrees of freedom reflect the fact that estimating σ with s costs one degree of freedom. As n increases, the t-distribution approaches the standard normal (z) distribution.

The test statistic measures how many standard errors the sample mean x̄ falls from the hypothesized value μ₀. A large absolute value of t indicates that the observed data are unlikely under H₀, which in turn provides evidence against the null. Notice that the formula uses s (sample standard deviation) rather than σ (population standard deviation), which is precisely why we need the t-distribution instead of the z-distribution—we are estimating the variability from the data itself. This distinction is critical: on the AP exam, if you are given σ (rare in practice), a z-test is appropriate; otherwise, a t-test is required.

💡

AP Exam Tip

When writing hypotheses on the AP exam, always define the parameter μ in context using a complete sentence before writing the symbolic form. For example: "Let μ = the true mean commute time (in minutes) for all employees at this company. H₀: μ = 25 and Hₐ: μ > 25." Omitting the contextual definition is a common reason students lose the 'identify' point on free-response questions.

SECTION 5

Choosing the Correct Tail Direction

One of the most consequential decisions in setting up a hypothesis test is choosing whether the alternative hypothesis is one-tailed (left or right) or two-tailed. This choice is driven entirely by the research question and must be made before the data are analyzed. Key language clues in the problem stem guide your decision, and the following diagram and table show exactly how to match common phrasings to the correct form of Hₐ.

The top row shows the three possible rejection regions (shaded areas) on a t-distribution. In a left-tailed test, only extreme negative t-values lead to rejection; in a right-tailed test, only extreme positive values; in a two-tailed test, extreme values in either direction lead to rejection, with the significance level α split equally between both tails. The bottom table lists common language clues that map to each direction.

A common source of confusion is the distinction between a two-tailed test and two separate one-tailed tests. A two-tailed test (Hₐ: μ ≠ μ₀) is appropriate when you want to detect a deviation in either direction—for instance, when a quality control engineer needs to know whether a machine is filling bottles with too much or too little liquid. A one-tailed test is appropriate when the research question focuses on a specific directional claim, such as a pharmaceutical company testing whether a new drug lowers blood pressure. Using a two-tailed test when the question is directional reduces statistical power, while using a one-tailed test to fish for significance after seeing the data is a violation of the testing framework.

SECTION 6

Worked Example

A city health department claims that the average systolic blood pressure of adults in the city is 120 mmHg. A researcher suspects the true mean may be higher due to local dietary habits. She obtains a simple random sample of 45 adults and records their systolic blood pressures, finding x̄ = 124.3 mmHg and s = 14.8 mmHg. Set up the appropriate hypothesis test.

Setting Up a One-Sample t-Test for Systolic Blood Pressure

1
Step 1 — Identify the Parameter

We define the parameter of interest in context. Let μ = the true mean systolic blood pressure (in mmHg) of all adults in the city. Note that μ refers to the entire population, not just the 45 sampled adults.

μ = true mean systolic blood pressure (mmHg) of all city adults
2
Step 2 — State the Hypotheses

The health department's claim (μ = 120) becomes the null hypothesis. Because the researcher "suspects the true mean may be higher," the alternative is one-sided to the right.

H₀: μ = 120 Hₐ: μ > 120
3
Step 3 — Identify the Test and Significance Level

Because we are testing a single population mean with σ unknown (we only have s = 14.8), the appropriate procedure is a one-sample t-test. Unless otherwise specified, we use α = 0.05.

Procedure: one-sample t-test at α = 0.05
4
Step 4 — Check Conditions

Random: The problem states the researcher used a simple random sample of 45 adults, so this condition is met. Independence (10% condition): It is reasonable to assume the city has at least 10 × 45 = 450 adults, so the sample is less than 10% of the population; observations are approximately independent. Normal/Large Sample: Since n = 45 ≥ 30, the Central Limit Theorem ensures the sampling distribution of x̄ is approximately normal regardless of the population shape.

All three conditions (Random, Independent, Normal/Large Sample) are satisfied ✓
5
Step 5 — Summary of Setup

The test is now fully set up. The next steps (which are beyond the scope of this lesson's focus) would involve computing the test statistic t = (124.3 − 120) / (14.8 / √45), finding the p-value from the t-distribution with df = 44, and drawing a conclusion in context.

Setup complete — ready to compute t-statistic and p-value

SECTION 7

Common Errors & How to Avoid Them

Free-response scoring guides from the College Board reveal several recurring errors that cost students points on hypothesis test setup questions. The table below catalogs the most common mistakes alongside the correct approach, so you can develop habits that translate directly into exam credit.

Common Error	Why It's Wrong	Correct Approach
Writing H₀: x̄ = 120	x̄ is a sample statistic, not a parameter. Hypotheses are about the population.	H₀: μ = 120. Always use the population parameter μ.
Choosing tail direction after seeing data	This inflates the Type I error rate and violates the test's validity.	Determine the direction from the research question before analysis.
Failing to define μ in context	Without context, the reader cannot verify the hypotheses match the problem.	"Let μ = the true mean [quantity] of [population], in [units]."
Writing H₀: μ ≥ 120 or H₀: μ ≤ 120	In AP Statistics, the null is always a statement of equality (μ = μ₀).	Use H₀: μ = μ₀. The inequality goes only in Hₐ.
Listing conditions without justification	Simply naming "Random, Normal, Independent" without connecting to the problem earns minimal credit.	Explain: "SRS is stated; n = 45 ≥ 30 so CLT applies; 45 < 10% of city population."

Common errors in hypothesis test setup and their corrections

✦

KEY TAKEAWAY

Think of the hypothesis setup as the blueprint for a building—if the blueprint contains errors, even perfect construction will yield a flawed structure. A misidentified parameter, a reversed tail direction, or an unchecked condition will invalidate every subsequent calculation. On the AP exam, the setup is typically worth 2–3 of the 4 points on an inference free-response question, so investing time in a careful, context-rich setup is the highest-return strategy.

SECTION 8

Connection to Confidence Intervals and Two-Sample Tests

The one-sample t-test for a population mean does not exist in isolation—it belongs to a broader family of inference procedures. Understanding how the test setup connects to confidence intervals and extends to two-sample scenarios will deepen your conceptual understanding and help you navigate the full scope of AP Statistics inference questions.

Feature	One-Sample t-Test (This Lesson)	Confidence Interval for μ	Two-Sample t-Test
Goal	Test whether μ equals a specific value μ₀	Estimate a range of plausible values for μ	Test whether μ₁ − μ₂ equals 0 (or another value)
Parameter	μ (one population mean)	μ (one population mean)	μ₁ − μ₂ (difference of two means)
Hypotheses	H₀: μ = μ₀; Hₐ: μ (<, >, ≠) μ₀	No formal hypotheses; instead, interpret interval	H₀: μ₁ − μ₂ = 0; Hₐ: μ₁ − μ₂ (<, >, ≠) 0
Conditions	Random, Independent, Normal/Large Sample	Same three conditions	Same three conditions, checked for each sample independently
Key Relationship	Reject H₀ if p-value < α	If μ₀ is not in the CI, a two-tailed test would reject H₀	Setup extends naturally; uses two x̄ values and pooled or unpooled SE

Comparison of one-sample t-test with related inference procedures

A particularly elegant connection exists between the one-sample t-test and the confidence interval. For a two-tailed test at significance level α, rejecting H₀ is equivalent to finding that μ₀ falls outside the (1 − α) × 100% confidence interval for μ. This duality means that if you construct a 95% confidence interval and the hypothesized value μ₀ lies outside it, a two-tailed test at α = 0.05 would also reject H₀. As you progress through the AP Statistics curriculum, you will encounter paired t-tests and two-sample t-tests; the setup logic you learn here—define the parameter, write hypotheses, check conditions—transfers directly to those more complex scenarios.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A researcher wants to test whether the mean amount of sleep college students get per night differs from the recommended 8 hours. Which of the following correctly states the null and alternative hypotheses?

A. H₀: x̄ = 8; Hₐ: x̄ ≠ 8
B. H₀: μ = 8; Hₐ: μ < 8
C. H₀: μ = 8; Hₐ: μ ≠ 8
D. H₀: μ ≠ 8; Hₐ: μ = 8
E. H₀: μ > 8; Hₐ: μ = 8

Show answer (C)

The correct answer is C. The word "differs" indicates a two-tailed test, so Hₐ uses ≠. Choice A is wrong because hypotheses must be about the population parameter μ, not the sample statistic x̄. Choice B is wrong because "differs" does not suggest a specific direction (less than); it suggests the mean could be either above or below 8. Choice D reverses the roles: the null must always be the equality statement. Choice E incorrectly places the inequality in H₀ and the equality in Hₐ.

Question 2: BASIC CALCULATION

A car manufacturer claims the mean fuel efficiency of a new sedan model is 35 miles per gallon. A consumer group believes the true mean is less than 35 mpg. They plan to take a random sample of 50 sedans and measure fuel efficiency. Which of the following is the correct alternative hypothesis?

A. Hₐ: μ > 35
B. Hₐ: μ ≠ 35
C. Hₐ: x̄ < 35
D. Hₐ: μ < 35
E. Hₐ: μ ≤ 35

Show answer (D)

The correct answer is D. The consumer group "believes the true mean is less than 35 mpg," which directly translates to Hₐ: μ < 35 (a left-tailed test). Choice A points in the wrong direction. Choice B is two-tailed, which does not match the specific directional claim. Choice C incorrectly uses the sample statistic x̄ instead of the population parameter μ. Choice E uses ≤, but the alternative hypothesis requires a strict inequality in the AP Statistics framework.

Question 3: INTERMEDIATE

A nutritionist selects a random sample of 22 protein bars from a production run and measures their calorie counts. The sample mean is 213 calories with a standard deviation of 18 calories. The label claims each bar contains 200 calories, and the nutritionist suspects the bars actually contain more. The distribution of calorie counts in the sample appears roughly symmetric with no outliers. Which of the following correctly identifies the appropriate test and explains why the Normal/Large Sample condition is satisfied?

A. One-sample z-test; the sample size of 22 is large enough by the CLT.
B. One-sample t-test; n = 22 ≥ 30 so the CLT applies.
C. One-sample t-test; although n = 22 < 30, the sample data are roughly symmetric with no outliers, so the Normal condition is reasonable.
D. One-sample t-test; the population of protein bars is normally distributed.
E. Two-sample t-test; we are comparing the sample to the labeled value.

Show answer (C)

The correct answer is C. Since σ is unknown (we only have s = 18), we use a t-test, not a z-test, eliminating choice A. Choice B incorrectly claims n = 22 ≥ 30; it does not. Choice C correctly identifies that with n < 30, we rely on the shape of the sample distribution—the problem states it is roughly symmetric with no outliers, which is sufficient to proceed. Choice D claims the population is normal, but the problem gives no information about the population distribution, only the sample. Choice E is incorrect because this is a one-sample scenario comparing one mean to a hypothesized value, not comparing two independent groups.

Question 4: APPLIED

4 pts

An environmental engineer measures dissolved oxygen levels (in mg/L) in a random sample of 36 water specimens from a lake. Historical records indicate that the mean dissolved oxygen level has been 7.5 mg/L. Recent industrial activity upstream leads the engineer to suspect the level has decreased. The sample yields x̄ = 7.1 mg/L and s = 1.2 mg/L. (a) Define the parameter of interest in context. (b) State the null and alternative hypotheses using appropriate notation. (c) Identify the appropriate inference procedure and justify your choice. (d) Verify all conditions required for the test, with justification for each.

Show answer

Point 1 — Parameter Definition (1 point): Let μ = the true mean dissolved oxygen level (in mg/L) in the lake. The parameter must be defined using μ (not x̄), must reference the population (the lake, not just the sample), and must include units or context. Point 2 — Hypotheses (1 point): H₀: μ = 7.5 and Hₐ: μ < 7.5. The null uses equality at the historical value; the alternative is left-tailed because the engineer suspects a decrease. Scoring note: using ≤ or ≥ in either hypothesis, using x̄ instead of μ, or using the wrong direction loses this point. Point 3 — Identify Procedure (1 point): The appropriate procedure is a one-sample t-test because we are testing a single population mean and σ is unknown (only s is available from the sample). A z-test would be inappropriate since the population standard deviation is not given. Point 4 — Conditions (1 point): (1) Random: The problem states the specimens were a random sample. (2) Independence (10% condition): 36 specimens is almost certainly less than 10% of all water specimens that could be drawn from the lake, so independence is reasonable. (3) Normal/Large Sample: n = 36 ≥ 30, so by the Central Limit Theorem, the sampling distribution of x̄ is approximately normal. All three conditions are satisfied. Scoring note: all three conditions must be named and justified with problem-specific details to earn this point.

Question 5: CRITICAL THINKING

4 pts

A pharmaceutical company is testing a new formulation of a pain-relief tablet. The old formulation has a known mean dissolution time of 12 minutes. The company hopes the new formulation dissolves faster, but wants to ensure it does not dissolve too slowly either. A biostatistician argues for a two-tailed test (Hₐ: μ ≠ 12), while a product manager argues for a one-tailed test (Hₐ: μ < 12). (a) Explain the reasoning behind each person's recommendation. (b) Discuss how the choice of tail direction affects the probability of committing a Type II error (failing to detect a real decrease in dissolution time). (c) Recommend which test the company should use, providing a clear statistical and practical justification. (d) Suppose the company proceeds with a one-tailed test at α = 0.05 and the sample data yield a p-value of 0.03. State the conclusion in context, and then explain what would change if a two-tailed test had been used with the same data.

Show answer

Point 1 — Reasoning for Each Recommendation (1 point): The biostatistician argues for a two-tailed test because, from a safety and regulatory perspective, the company needs to detect any change in dissolution time—whether faster or slower—to ensure the new formulation performs acceptably. The product manager argues for a one-tailed test because the company's primary research goal is to demonstrate that the new formulation dissolves faster; slower dissolution would simply mean the new formulation is no better than the old, and the company would not adopt it regardless. Point 2 — Type II Error Discussion (1 point): For a fixed significance level α, a one-tailed test has more power (lower probability of Type II error) to detect a difference in the specified direction than a two-tailed test. This is because the entire α is allocated to one tail, making the critical value less extreme. If the true mean dissolution time is, say, 11 minutes (genuinely faster), the one-tailed test is more likely to reject H₀ and detect this improvement. The two-tailed test splits α across both tails, requiring a more extreme test statistic to reject H₀ in either direction, thus increasing the chance of a Type II error for a directional effect. Point 3 — Recommendation with Justification (1 point): The most defensible choice depends on context. If regulatory bodies require the company to demonstrate that the formulation has not changed dissolution time in a harmful way (i.e., too slow), a two-tailed test is appropriate. However, if the sole research question is whether the new formulation is faster, a one-tailed test is appropriate and offers greater power. In practice, the company likely cares about both safety and efficacy, so a two-tailed test is the more conservative and defensible choice—it protects against undetected increases in dissolution time. Any well-justified recommendation earns credit. Point 4 — Conclusion and Comparison (1 point): With a one-tailed test at α = 0.05 and a p-value of 0.03, we reject H₀ because 0.03 < 0.05. In context: there is convincing evidence that the true mean dissolution time of the new formulation is less than 12 minutes. If a two-tailed test had been used with the same data, the p-value would be approximately 2 × 0.03 = 0.06, which exceeds 0.05. We would fail to reject H₀ and conclude that there is not convincing evidence of a difference. This illustrates how the choice of tail direction can change the conclusion even with identical data, reinforcing the importance of choosing the direction before data collection.

SUMMARY

Lesson Summary

Summary

Setting up a hypothesis test for a population mean requires four essential steps. First, define the parameter μ in context, specifying the population and units. Second, write the null hypothesis H₀: μ = μ₀ (always an equality) and the alternative hypothesis Hₐ with the direction (<, >, or ≠) determined by the research question, not the data. Third, identify the procedure as a one-sample t-test (since σ is almost always unknown). Fourth, verify the three conditions: Random (data from a random sample or experiment), Independence (10% condition: n ≤ 0.10N), and Normal/Large Sample (population approximately normal, or n ≥ 30 by CLT, or sample data show no strong skewness or outliers for smaller samples).

Remember that the tail direction must be chosen before examining the data, and hypotheses are always stated in terms of the population parameter μ, never the sample statistic x̄. On the AP exam, the setup is typically worth the majority of the points on inference free-response questions, so a careful, context-rich setup is your highest-return investment.