Hypotheses Describe Population Means
A two-sample t test begins with a clear statement of what the two populations are and what difference between them is of interest. In “Comparing Two Means: The Parameter of Interest,” you saw that \(\mu_1-\mu_2\) represents the true mean for population 1 minus the true mean for population 2. For a hypothesis test, the null hypothesis describes a reference claim about that population difference, and the alternative describes the kind of difference the investigation is designed to detect.
The hypotheses are about population parameters, not about the sample means \(\bar{x}_1\) and \(\bar{x}_2\). The sample means provide evidence when a test is carried out; they do not define the hypotheses.
The population labels matter because they establish the subtraction order. If population 1 is the group using a new method and population 2 is the group using a standard method, then \(\mu_1-\mu_2\) means new method minus standard method. Reversing the labels reverses the sign of the difference.
The Null Hypothesis: No Difference
For a test asking whether two population means differ, the null hypothesis states that the population means are equal. This can be written as \(H_0:\mu_1=\mu_2\), or equivalently as \(H_0:\mu_1-\mu_2=0\). The second form makes the parameter being tested especially visible: zero is the value that represents no difference between the means.
The equality belongs in the null hypothesis. It describes the reference value against which the observed difference in sample means will later be assessed. Although the sample means may not be exactly equal, the null hypothesis is not a claim that the sample means are equal.
For example, if \(\mu_1\) is the true mean battery life for devices made by Company A and \(\mu_2\) is the true mean battery life for devices made by Company B, then \(H_0:\mu_1-\mu_2=0\) says that the two companies’ population mean battery lives are the same. The parameter is measured in hours.
Alternative Hypotheses Describe the Research Claim
The alternative hypothesis, \(H_a\), describes the difference the study is set up to find evidence for. It must use one of three forms: not equal to zero, greater than zero, or less than zero. The symbols apply to the defined population means, in the stated subtraction order.
A two-sided alternative, \(H_a:\mu_1-\mu_2\ne0\), represents a claim that the means differ, without specifying which population mean is larger. A one-sided alternative specifies a direction. If \(H_a:\mu_1-\mu_2>0\), then \(\mu_1\) is greater than \(\mu_2\); if \(H_a:\mu_1-\mu_2<0\), then \(\mu_1\) is less than \(\mu_2\).
The direction must match the meaning of the subtraction. For example, \(H_a:\mu_1-\mu_2>0\) does not mean that both means are positive. It says that the first population’s mean is greater than the second population’s mean. “Choosing One-Sided or Two-Sided for Two Means,” the next tutorial, explores how to match an alternative to a research question. Here, the essential skill is to write the hypotheses correctly once the populations, variable, and research claim are defined.
A Reliable Way to Write the Hypotheses
Before writing symbols, identify the quantitative variable and the two populations. Then assign population 1 and population 2, explicitly state what each mean represents, and decide what claim the study is investigating. Translate that claim into a comparison of \(\mu_1\) and \(\mu_2\), preserving the chosen order.
State what \(\mu_1\) and \(\mu_2\) mean, including the populations and the quantitative variable.
Represent no difference with \(H_0:\mu_1=\mu_2\) or \(H_0:\mu_1-\mu_2=0\).
Use \(\ne\) for a claim of a difference in either direction, \(>\) for a positive difference in the stated order, or \(<\) for a negative difference.
Confirm that reversing the populations would reverse the sign, and identify the units for the mean difference.
This is a setup, not a calculation. Do not choose the hypotheses by looking first at which sample mean is larger. The hypotheses should reflect the research question and study plan, rather than a pattern noticed after collecting data.
Worked Examples
Worked Example: Comparing Two Water Filters
An invented laboratory comparison measures the volume of water, in liters, that two filter designs process before replacement. Researchers want to know whether the designs have different mean capacities. Write the hypotheses.
Define the parameters: Let \(\mu_1\) be the true mean capacity, in liters, for all filters of Design A under the specified test conditions. Let \(\mu_2\) be the true mean capacity, in liters, for all filters of Design B under those conditions.
Write the null: No difference in mean capacity means \(\mu_1-\mu_2=0\), so \(H_0:\mu_1-\mu_2=0\). Equivalently, \(H_0:\mu_1=\mu_2\).
Write the alternative: The research question asks whether the designs have different mean capacities but does not specify which should be higher. The alternative is \(H_a:\mu_1-\mu_2\ne0\).
Check the interpretation: The parameter difference is measured in liters. A positive value would mean Design A has the higher population mean capacity; a negative value would mean Design B has the higher population mean. The alternative allows either direction.
Worked Example: A New Schedule and Sleep Duration
In an invented school study, students are assigned to one of two schedule formats. Researchers expect students assigned to Schedule A to average more hours of sleep per night than students assigned to Schedule B. Write the hypotheses for comparing the two population means.
Define the parameters: Let \(\mu_A\) be the true mean hours of sleep per night for all students who would follow Schedule A. Let \(\mu_B\) be the true mean hours of sleep per night for all students who would follow Schedule B.
Set the subtraction order: Use Schedule A first and Schedule B second. The difference of interest is \(\mu_A-\mu_B\), measured in hours of sleep per night.
Write the null: The no-difference reference claim is \(H_0:\mu_A-\mu_B=0\), equivalently \(H_0:\mu_A=\mu_B\).
Write the alternative: The expectation is that Schedule A has the larger population mean, so the alternative is \(H_a:\mu_A-\mu_B>0\).
Check the interpretation: A positive difference means the mean sleep duration for Schedule A is greater than the mean for Schedule B. It does not claim that every student following Schedule A sleeps more than every student following Schedule B.
Worked Example: Comparing Two Delivery Methods
An invented logistics team compares the time, in minutes, required to deliver a standard package using Route X and Route Y. The team’s specific question is whether Route X has a lower true mean delivery time than Route Y. Write the hypotheses.
Define the parameters: Let \(\mu_X\) be the true mean delivery time for all comparable packages delivered using Route X under the stated operating conditions. Let \(\mu_Y\) be the true mean delivery time for comparable packages delivered using Route Y under those conditions.
Set the subtraction order: Define the difference as Route X minus Route Y: \(\mu_X-\mu_Y\). Its units are minutes.
Write the null: The reference claim of no difference is \(H_0:\mu_X-\mu_Y=0\), or \(H_0:\mu_X=\mu_Y\).
Write the alternative: “Route X has a lower mean time than Route Y” translates to \(\mu_X<\mu_Y\), which is \(H_a:\mu_X-\mu_Y<0\).
Check the sign: Because the first mean is Route X’s and the second is Route Y’s, a negative difference means Route X’s mean delivery time is lower. If the order were reversed, the equivalent alternative would be \(\mu_Y-\mu_X>0\).
Worked Example: Keeping the Difference Order Straight
An invented environmental project compares the mean concentration of a substance in water samples from an upstream location and a downstream location. Let population 1 be the upstream location and population 2 the downstream location. The question is whether the mean concentrations differ. Write the hypotheses and explain the sign convention.
Define the parameters: Let \(\mu_1\) be the true mean concentration, in milligrams per liter, for water at the upstream location. Let \(\mu_2\) be the true mean concentration, in milligrams per liter, for water at the downstream location.
Write the null: No difference in population mean concentration is \(H_0:\mu_1-\mu_2=0\), equivalently \(H_0:\mu_1=\mu_2\).
Write the alternative: The question asks whether the locations differ in either direction, so \(H_a:\mu_1-\mu_2\ne0\).
Interpret the signs: A positive value of \(\mu_1-\mu_2\) means the upstream mean concentration is higher; a negative value means the downstream mean concentration is higher. Since the alternative is not equal to zero, it includes both possibilities.
Common Mistakes and AP Exam Tips
- Writing hypotheses about sample means: \(H_0:\bar{x}_1=\bar{x}_2\) is not the population claim being tested. Define \(\mu_1\) and \(\mu_2\), then write hypotheses using those parameters.
- Leaving out the population definitions: “Let \(\mu_1\) and \(\mu_2\) be the means” does not tell the reader which groups or variable they describe. Identify both populations and the quantitative variable in context.
- Putting equality in the alternative: A claim of a difference uses \(H_a:\mu_1-\mu_2\ne0\), not \(H_a:\mu_1-\mu_2=0\). The equality describes the null reference value.
- Using a direction that conflicts with the subtraction order: If \(\mu_1\) is Route X and \(\mu_2\) is Route Y, “Route X has a lower mean” requires \(\mu_1-\mu_2<0\). Explain the order before assigning the sign.
- Choosing an alternative after seeing the sample means: The hypotheses should reflect the research question, not whichever direction the observed sample difference happens to point. The sample results are evidence to evaluate later.
- Claiming the null is proved: The null is the reference claim for a test, not a conclusion guaranteed by writing it. A later test may provide convincing evidence against it, or the evidence may not be strong enough to reject it.
For full-credit communication, define both means in context, state the difference in the intended order, and write one equality null and one appropriate alternative. If you use \(H_0:\mu_1=\mu_2\), the corresponding alternatives are \(H_a:\mu_1\ne\mu_2\), \(H_a:\mu_1>\mu_2\), or \(H_a:\mu_1<\mu_2\), depending on the research claim.
Check Your Understanding
For each question, define the population means and write hypotheses that match the stated claim.
- A comparison asks whether two brands of rechargeable batteries have different mean operating times. Define the means and write \(H_0\) and \(H_a\).
- Let \(\mu_1\) be the true mean commute time for people using a bus and \(\mu_2\) the true mean for people using a train. Write the alternative for the claim that bus users have a lower mean commute time.
- For \(H_a:\mu_1-\mu_2>0\), explain in context what a positive difference means, assuming \(\mu_1\) and \(\mu_2\) are the means for two defined populations.
- Why should hypotheses be written using \(\mu_1\) and \(\mu_2\), rather than \(\bar{x}_1\) and \(\bar{x}_2\)?
- If a study instead defines its parameter as \(\mu_2-\mu_1\), how does the alternative equivalent to \(H_a:\mu_1-\mu_2<0\) change?