One-Sided Versus Two-Sided Alternatives for a Mean

The Question Determines the Direction

In “Stating Hypotheses for a Population Mean,” you learned to define $\mu$ in context, write the null hypothesis at a reference value, and use the alternative to express the question of interest. This tutorial focuses on one choice within that process: whether the alternative points in one direction or allows departures in either direction.

A test with a one-sided alternative looks for evidence that the population mean is specifically lower or specifically higher than the reference value. A test with a two-sided alternative looks for evidence that the population mean differs from the reference value in either direction. The wording of the claim or the concern being investigated should determine which question is being asked.

Definition: For a reference value $\mu_0$, a lower one-sided alternative is $H_a:\mu<\mu_0$, a higher one-sided alternative is $H_a:\mu>\mu_0$, and a two-sided alternative is $H_a:\mu\ne\mu_0$. Each states a different kind of departure from the null value.

For instance, “the mean is less than 12” points to $H_a:\mu<12$. “The mean is greater than 12” points to $H_a:\mu>12$. “The mean differs from 12” allows either a lower or a higher value, so it points to $H_a:\mu\ne12$. Match the direction to what the question treats as important—not to what the sample happens to show.

The null hypothesis continues to use equality at the reference value: $H_0:\mu=\mu_0$. This tutorial does not calculate a test statistic or decide whether evidence is convincing. Its goal is to state the question correctly before using sample data to assess it.

A Practical Way to Choose

Before writing symbols, identify the words that express the research question. Ask whether the question specifically asks about a lower mean, a higher mean, or any difference. If it does not specify a direction, do not invent one. A directional alternative is justified when a claim, a stated concern, or the purpose of the investigation identifies the relevant direction.

Define the parameter.
State what population mean $\mu$ describes, including the population and the quantitative variable. Keep the context and units clear.

Identify the reference value.
Find the benchmark or claimed mean, $\mu_0$, that appears in the question.

Translate the question’s direction.
“Lower” or “less than” points to $H_a:\mu<\mu_0$; “higher” or “greater than” points to $H_a:\mu>\mu_0$; “different” or “has changed” points to $H_a:\mu\ne\mu_0$, unless the wording specifies a direction.

Write the null and alternative together.
Use $H_0:\mu=\mu_0$, then check that $H_a$ expresses the question—not a guess based on the sample result.

“Up,” “down,” “increase,” “decrease,” “less,” and “more” often signal a direction, but read the whole sentence. “Has the mean changed?” asks about either direction. “Has the mean decreased?” asks about only a decrease. A number in a sentence may be a benchmark, but it does not by itself tell you whether the alternative is one-sided or two-sided.

Decision check: Imagine that the true mean were on the opposite side of the reference value from the one you are considering. Would that result answer the stated research question? If either direction matters, use a two-sided alternative. If the question is specifically about one direction, use the corresponding one-sided alternative.

Worked Examples

Worked Example: A Lower Battery-Life Concern

A fictional device maker states that the mean battery life of a particular model is 10 hours. A technician suspects that a recent software update has reduced average battery life. A random sample of 36 devices has a mean battery life of 10.3 hours. State hypotheses for the technician’s question.

Define the parameter. Let $\mu$ be the mean battery life, in hours, for all devices of this model running the recent software update.

Identify the direction. The stated suspicion is that the update has reduced battery life. “Reduced” means lower than the 10-hour benchmark, so the question is specifically about a lower population mean.

Write the hypotheses. $$ H_0:\mu=10\text{ hours} \qquad\text{and}\qquad H_a:\mu<10\text{ hours}. $$

Check the choice. The sample mean, 10.3 hours, is above 10 hours, but it does not change the hypotheses. The research question was set by the stated suspicion. The sample mean points away from the lower-mean alternative. A test using the sample variability would assess the strength of evidence against the null.

Worked Example: Checking Whether a Filling Mean Differs

A fictional juice company labels its cartons as containing an average of 750 mL. A quality team randomly samples 30 cartons from a production period. The team wants to find out whether the actual mean fill volume differs from 750 mL; it has not identified underfilling or overfilling as the particular concern.

Define the parameter. Let $\mu$ be the mean fill volume, in milliliters, of all cartons produced during the specified period.

Identify the direction. The team asks whether the mean “differs” from 750 mL. A mean below 750 and a mean above 750 would both answer that question, so the alternative must allow either direction.

Write the hypotheses. $$ H_0:\mu=750\text{ mL} \qquad\text{and}\qquad H_a:\mu\ne750\text{ mL}. $$

Explain why this is two-sided. The team has not limited its concern to a mean that is too low. Choosing $H_a:\mu<750$ just because a sample mean happened to fall below 750 would change the question after seeing the data. The stated question calls for a two-sided alternative.

Worked Example: Testing for a Longer Commute Time

A school transportation coordinator has historically used 24 minutes as a planning benchmark for the mean bus commute. After routes were reorganized, the coordinator suspects that mean commute time has increased. A random sample of 42 student commutes has a mean of 25.1 minutes. Write hypotheses that match the suspicion.

Define the parameter. Let $\mu$ be the mean commute time, in minutes, for all students riding the school buses on the reorganized routes during the period of interest.

Identify the direction. The coordinator suspects an increase, meaning a mean greater than the 24-minute benchmark. A possible decrease is not the concern named in this question.

Write the hypotheses. $$ H_0:\mu=24\text{ minutes} \qquad\text{and}\qquad H_a:\mu>24\text{ minutes}. $$

Check the choice. The sample mean of 25.1 minutes is above 24, which happens to point in the same direction as the suspicion. The alternative is still selected from the research question, not from that agreement. If the question instead asked whether routes had changed the mean commute time in either direction, the alternative would be $H_a:\mu\ne24$ minutes.

Worked Example: Clarifying an Ambiguous “Problem”

A fictional greenhouse uses 18°C as its target mean overnight temperature. A manager says, “I think we have a temperature problem,” and asks an analyst to test whether the mean differs from the target. A sample of 32 nights has a mean of 17.6°C. What alternative matches the explicit question?

Define the parameter. Let $\mu$ be the mean overnight temperature, in degrees Celsius, in the greenhouse during the specified operating period.

Separate the vague concern from the test question. “A temperature problem” is not precise about direction: temperatures that are too low and temperatures that are too high could both be problematic. More importantly, the manager’s explicit question is whether the mean differs from the target.

Write the hypotheses. $$ H_0:\mu=18^\circ\text{C} \qquad\text{and}\qquad H_a:\mu\ne18^\circ\text{C}. $$

Explain what would change the choice. If the manager had said that the concern was specifically that the greenhouse was too cold, the directional alternative would be $H_a:\mu<18^\circ\text{C}$. The observed sample mean of 17.6°C does not, by itself, justify changing the stated two-sided question into a lower-tail question.

Common Mistakes and AP Exam Tips

Choosing the alternative is not a matter of finding the symbol that matches the sample mean. It is a matter of accurately recording the question to be investigated. The distinction matters because $H_a:\mu<\mu_0$, $H_a:\mu>\mu_0$, and $H_a:\mu\ne\mu_0$ describe different claims.

Choosing the direction after seeing the sample mean. A sample mean below the benchmark does not automatically call for a lower alternative. A full-credit response uses the direction specified by the claim or suspicion, even if the sample mean points the other way.
Using a one-sided alternative for “different.” “Different,” “not equal,” and “has changed” allow departures in both directions when no direction is specified. State $H_a:\mu\ne\mu_0$, not just “less than” or “greater than.”
Using a two-sided alternative when the question is explicitly directional. If the question asks whether a mean increased, $H_a:\mu>\mu_0$ matches it. A two-sided alternative asks a broader question that also counts a decrease as a departure of interest.
Assuming “concern” always means one-sided. A concern may be about either direction. If the prompt does not say which direction matters, look for a more specific research question or ask for clarification instead of guessing.
Putting a directional inequality in the null. For the one-sample mean test in this course, state the null at the reference value using equality, such as $H_0:\mu=24$. Put the direction of interest in $H_a$.
Leaving the parameter undefined. Symbols alone may not show what population or variable is involved. Define $\mu$ in context and include the relevant units, as in the examples.

A strong response connects each symbol to the prompt: define $\mu$, identify the benchmark, quote or paraphrase the directional wording, and write the null and alternative. If the wording supports either direction, say so and use a two-sided alternative. If it names a particular direction, identify that direction directly.

Key takeaway: Choose $H_a:\mu<\mu_0$ for a specifically lower mean, $H_a:\mu>\mu_0$ for a specifically higher mean, and $H_a:\mu\ne\mu_0$ when departures in either direction matter. Let the claim or research question—not the observed sample mean—determine the direction.

Check Your Understanding

For each situation, define the population mean in context and write hypotheses that match the stated question. Explain briefly why the alternative is one-sided or two-sided.

A fictional orchard’s target mean apple mass is 160 g. The manager wants to know whether the mean mass of this week’s apples is below the target.
A school sets a 25-minute planning benchmark for student travel time. Administrators ask whether the mean travel time differs from 25 minutes, without naming a direction.
A fictional water-treatment team’s benchmark mean processing time is 12 minutes. The team suspects that a new procedure has increased the mean time.
A sample of 28 headphones has a mean battery life of 7.8 hours. The advertised benchmark is 8 hours, and the question asks whether the mean battery life has changed. Does the sample mean determine the direction?
A manager says that the mean temperature is “not right,” but does not say whether it is too high or too low. What information is needed to choose between a one-sided and a two-sided alternative?