01 / 21

Setting Up Tests for
Population Mean

Learn how to formulate null and alternative hypotheses, identify the correct test statistic, and verify conditions for a one-sample \( t \)-test on \( \mu \).

AP Statistics

What Is a Hypothesis Test?

A hypothesis test is a formal procedure that uses sample data to evaluate a claim about a population parameter. For this unit, the parameter of interest is the population mean \( \mu \).

Claim
A statement about \( \mu \), such as "the average wait time is 10 minutes."
Evidence
Sample data: \( \bar{x} \), \( s \), and \( n \) collected from the real world.
Decision
Reject or fail to reject the claim based on how unlikely the data are.

The Two Hypotheses

Every test begins by writing two competing statements about \( \mu \):

\( H_0 \): Null Hypothesis
The "no change" or "no difference" claim. Always states equality: \( H_0\colon \mu = \mu_0 \).
\( H_a \): Alternative Hypothesis
The research claim. States the direction of the difference: \( \mu \neq \mu_0 \), \( \mu < \mu_0 \), or \( \mu > \mu_0 \).

Why Does Setting Up Matter?

An incorrect setup leads to an entirely wrong conclusion, even if every calculation afterward is perfect. Getting the hypotheses, test statistic, and conditions right is the foundation of any valid test.

State
\( H_0 \) and \( H_a \)
Check Conditions Compute & Conclude

Formal Hypothesis Structure

The null hypothesis always contains an equality. The alternative hypothesis indicates the direction of the test.

1
Two-tailed: \( H_0\colon \mu = \mu_0 \) vs. \( H_a\colon \mu \neq \mu_0 \)
2
Left-tailed: \( H_0\colon \mu = \mu_0 \) vs. \( H_a\colon \mu < \mu_0 \)
3
Right-tailed: \( H_0\colon \mu = \mu_0 \) vs. \( H_a\colon \mu > \mu_0 \)

Choosing the Direction of \( H_a \)

The wording of the research question tells you which alternative to use. Look for key phrases:

Key Phrase \( H_a \) Form Tail
"is different from," "has changed" \( \mu \neq \mu_0 \) Two
"is less than," "has decreased" \( \mu < \mu_0 \) Left
"is greater than," "has increased" \( \mu > \mu_0 \) Right

The Test Statistic: One-Sample \( t \)

When \( \sigma \) is unknown (which is almost always), we use the one-sample \( t \)-statistic:

\( \bar{x} \)
Sample mean
\( \mu_0 \)
Hypothesized population mean
\( s \)
Sample standard deviation
\( n \)
Sample size

\( t \)-test vs. \( z \)-test: Key Distinction

Feature \( z \)-test \( t \)-test
\( \sigma \) known? Yes No (use \( s \))
Distribution Standard Normal \( t \) with \( df = n - 1 \)
AP Stats usage Rare (proportions) Primary for means
Tails thicker? No Yes, especially small \( n \)

The Three Conditions

Before computing the test statistic, you must verify three conditions. AP graders look for all three.

1
Random: Data come from a random sample or randomized experiment.
2
Independence (10% Rule): When sampling without replacement, the sample size \( n \) must be < 10% of the population \( N \).
3
Normal/Large Sample: The population is approximately Normal, or \( n \geq 30 \) (CLT), or a graph of the data shows no strong skewness or outliers.

Checking the Normal Condition

The guideline depends on sample size. Smaller samples require stronger evidence of normality.

\( n < 15 \)
Must look
roughly Normal
No outliers or
strong skew
\( 15 \leq n < 30 \)
Moderate skew OK
if no outliers
Check dotplot or
boxplot
\( n \geq 30 \)
CLT applies
Shape of population
matters less

How to Set Up the Test

1
Define the parameter: "Let \( \mu \) = the true mean [context]."
2
State hypotheses: Write \( H_0\colon \mu = \mu_0 \) and the appropriate \( H_a \).
3
Choose significance level: Usually \( \alpha = 0.05 \), unless told otherwise.
4
Name the test: "One-sample \( t \)-test for \( \mu \)."
5
Verify conditions: Random, 10% rule, Normal/Large Sample.

Example 1: Coffee Shop Wait

A coffee shop claims the average wait time is 5 minutes. A random sample of \( n = 40 \) customers has \( \bar{x} = 5.8 \) min and \( s = 2.1 \) min. Is there evidence the true mean wait is greater than 5 minutes?

1
Parameter: Let \( \mu \) = true mean wait time (min) for all customers.
2
Hypotheses: \( H_0\colon \mu = 5 \) vs. \( H_a\colon \mu > 5 \) (right-tailed).
3
Test: One-sample \( t \)-test for \( \mu \) at \( \alpha = 0.05 \).

Example 1: Checking Conditions

✓
Random: The problem states a random sample was taken.
✓
10% Rule: \( n = 40 \) is < 10% of all customers (safe to assume thousands of customers).
✓
Normal/Large Sample: \( n = 40 \geq 30 \), so the CLT guarantees approximate normality of \( \bar{x} \).

Example 2: Battery Life

A manufacturer states its batteries last 500 hours on average. A consumer group randomly selects \( n = 18 \) batteries and finds \( \bar{x} = 486 \) hours, \( s = 24 \) hours. The dotplot shows a roughly symmetric distribution with no outliers. Is there evidence the true mean is different from 500?

1
Parameter: Let \( \mu \) = true mean lifetime (hours) of all batteries.
2
Hypotheses: \( H_0\colon \mu = 500 \) vs. \( H_a\colon \mu \neq 500 \) (two-tailed).
3
Conditions: Random ✓. \( 18 < 10\% \) of all batteries ✓. \( 15 \leq n < 30 \): dotplot shows no outliers and roughly symmetric ✓.

Example 3: Tricky Wording

A school principal believes students now sleep less than the recommended 8 hours. She surveys \( n = 50 \) randomly selected students and finds \( \bar{x} = 7.4 \) hours, \( s = 1.2 \) hours.

✓
Random: Randomly selected students.
✓
10% Rule: \( 50 < 10\% \) of all students in the school (assuming at least 500).
✓
Normal: \( n = 50 \geq 30 \), so CLT applies.

Common Mistakes to Avoid

✕ Using \( \bar{x} \) in hypotheses
Hypotheses are about the parameter \( \mu \), never the sample statistic \( \bar{x} \).
✕ Inequality in \( H_0 \)
The null must state equality: \( H_0\colon \mu = \mu_0 \). Writing \( H_0\colon \mu \leq 5 \) is incorrect on the AP exam.
✕ Skipping conditions
Failing to check Random, 10%, and Normal/Large Sample will cost you points on FRQs.

Question 1 of 5

A researcher claims the mean height of a plant species is different from 30 cm. Which hypotheses are correct?

Question 2 of 5

A sample of \( n = 12 \) has a heavily right-skewed distribution with one outlier. Which condition is not met for a one-sample \( t \)-test?

Question 3 of 5

A dietitian claims the average calorie intake for a certain group is less than 2000 calories. A random sample of \( n = 35 \) gives \( \bar{x} = 1940 \) and \( s = 180 \). What is the correct test statistic formula?

Question 4 of 5

A poll surveys 200 randomly chosen residents of a city with 5,000 residents. The 10% condition requires:

Question 5 of 5

A gym advertises that its members lose an average of 10 pounds in a month. A skeptical member believes the true average loss is less than 10 pounds. What are the correct hypotheses?