Null & Alternative Hypotheses - College Statistics

A dietitian wants to assess if a particular 4-week meal plan results in a change in cholesterol levels. The dietitian measures the cholesterol level of 30 subjects before the plan starts and again after the plan is completed. Let $\mu_d$ represent the true mean of the differences in cholesterol levels, calculated as (level $_{before}$ - level $_{after}$ ). To find evidence of a reduction in cholesterol, what hypotheses should be tested?
Answer: $H_0: \mu_d \le 0$ and $H_a: \mu_d > 0$ This is a paired-sample design, so the analysis should focus on the mean of the differences, $\mu_d$ . A reduction in cholesterol means the 'after' level is lower than the 'before' level. Since the difference is defined as $d = \text{level}_{before} - \text{level}_{after}$ , a reduction in cholesterol corresponds to a positive difference. The dietitian is looking for evidence of a reduction, so the alternative hypothesis is $H_a: \mu_d > 0$ . The null hypothesis is that there is no reduction or an increase, $H_0: \mu_d \le 0$ .
A school district implemented a new math curriculum for its 4th-grade students. To evaluate its effectiveness, they will compare the mean score on a standardized math test for students under the new curriculum ( $\mu_N$ ) with the mean score for students in a neighboring district that uses the old curriculum ( $\mu_O$ ). The district's goal is to show that its students now score higher. Which of the following represents the alternative hypothesis ( $H_a$ ) for this study?
Answer: $H_a: \mu_N > \mu_O$ When you encounter hypothesis testing questions, you need to carefully distinguish between what the researchers want to prove (the alternative hypothesis) and what they're testing against (the null hypothesis). The alternative hypothesis should directly reflect the research goal using population parameters, not sample statistics. The district wants to demonstrate that their new curriculum helps students score higher than the old curriculum. This means they're trying to prove that the population mean under the new curriculum ( $\mu_N$ ) is greater than the population mean under the old curriculum ( $\mu_O$ ). This research goal translates directly into the alternative hypothesis: $H_a: \mu_N > \mu_O$ . Choice A ( $H_a: \mu_N < \mu_O$ ) suggests the new curriculum is worse than the old one, which contradicts the district's goal of showing improvement. Choice B ( $H_a: \mu_N \neq \mu_O$ ) represents a two-tailed test that would detect any difference between curricula, but the district specifically wants to show their curriculum is better, not just different. Choice C ( $H_a: \bar{x}_N > \bar{x}_O$ ) makes a critical error by using sample means ( $\bar{x}$ ) instead of population means ( $\mu$ ). Hypotheses are always about population parameters, not sample statistics. Remember this key principle: the alternative hypothesis should match the researcher's specific claim using population parameters. If they want to show an increase, use "greater than" ( $>$ ). If they want to show a decrease, use "less than" ( $<$ ). If they just want to show a difference, use "not equal" ( $\neq$ ). Always use Greek letters for population parameters in formal hypotheses.
A company claims that the mean lifetime of its light bulbs is at least 1,200 hours. A consumer advocacy group suspects this claim is an overstatement and that the true mean lifetime is less. The group tests a random sample of these bulbs and finds a sample mean lifetime of 1,180 hours. What are the appropriate null and alternative hypotheses for the consumer group's test?
Answer: $H_0: \mu \ge 1200$ and $H_a: \mu < 1200$ Hypotheses are always stated about population parameters (like $\mu$ ), not sample statistics (like $\bar{x}$ ). The value of 1,180 hours is the sample mean, which is evidence, not part of the hypothesis itself. The consumer group is testing the company's claim. The group's suspicion is that the mean lifetime is less than 1,200 hours, which is the alternative hypothesis ( $H_a: \mu < 1200$ ). The null hypothesis is the company's claim that is being challenged, $H_0: \mu \ge 1200$ .
A machine is designed to fill bags of coffee to a mean weight of 340 grams. To ensure consistency and avoid financial losses from overfilling or legal issues from underfilling, a quality control manager needs to test if the machine is deviating from this target. The manager is concerned about any deviation, either high or low. Which hypotheses are appropriate for this quality control test?
Answer: $H_0: \mu = 340$ and $H_a: \mu \neq 340$ The manager is concerned about any deviation from the target of 340g, meaning the process is out of control if the mean weight is either significantly more or significantly less than 340g. This situation calls for a two-tailed test. The null hypothesis is that the machine is working as intended ( $H_0: \mu = 340$ ). The alternative hypothesis is that the machine's mean fill weight is not equal to the target ( $H_a: \mu \neq 340$ ).
An economist is using simple linear regression to study the effect of interest rates (x) on the monthly number of new housing starts (y). The economist wants to determine if there is sufficient evidence to conclude that higher interest rates are associated with a decrease in housing starts. Let $\beta_1$ represent the slope of the population regression line. What are the null and alternative hypotheses for this study?
Answer: $H_0: \beta_1 \ge 0$ and $H_a: \beta_1 < 0$ The economist is testing for a specific directional relationship: that higher interest rates lead to a decrease in housing starts. In a regression context, this corresponds to a negative slope. The research hypothesis is therefore that the population slope $\beta_1$ is less than 0. This becomes the alternative hypothesis ( $H_a: \beta_1 < 0$ ). The null hypothesis states that there is no negative relationship (i.e., the slope is zero or positive), so $H_0: \beta_1 \ge 0$ .
A new, less expensive manufacturing process for a medical component is being evaluated. The critical characteristic is its strength. The new process will be adopted if it is not inferior to the old process. Inferiority is defined as the mean strength of new components ( $\mu_{new}$ ) being more than 2 units below the mean strength of old components ( $\mu_{old}$ ). To gain regulatory approval, the manufacturer must present statistical evidence that the new process is not inferior. Which hypotheses should the manufacturer test?
Answer: $H_0: \mu_{new} - \mu_{old} \le -2$ and $H_a: \mu_{new} - \mu_{old} > -2$ This is a non-inferiority test. The burden of proof is on demonstrating that the new process is not inferior. The 'undesirable' outcome, or the state of inferiority, is that the new mean strength is more than 2 units below the old mean, which translates to $\mu_{new} < \mu_{old} - 2$ , or $\mu_{new} - \mu_{old} < -2$ . In hypothesis testing, we seek to reject the null hypothesis. Therefore, the null hypothesis is set as the state of inferiority, $H_0: \mu_{new} - \mu_{old} \le -2$ . The alternative hypothesis, which represents the conclusion the manufacturer wishes to support (non-inferiority), is $H_a: \mu_{new} - \mu_{old} > -2$ .
A market researcher is investigating whether there is an association between a consumer's age group (18-34, 35-54, 55+) and their primary method of grocery shopping (in-store, online pickup, online delivery). What are the null and alternative hypotheses for the most appropriate statistical test?
Answer: $H_0:$ There is no association between age group and primary shopping method. $H_a:$ There is an association between age group and primary shopping method. The researcher is examining the relationship between two categorical variables (age group and shopping method). The appropriate test for this is a chi-squared test of independence. The null hypothesis for this test is that the two variables are independent (i.e., there is no association between them). The alternative hypothesis is that the variables are dependent (i.e., there is an association).
An automotive engineer is designing a piston ring. The manufacturing process is specified to produce rings with a diameter variance of no more than 0.03 mm $^2$ . The engineer suspects that a recent adjustment to the machinery has increased this variance, making the process less precise. A sample of rings is collected to test this suspicion. Let $\sigma^2$ be the true variance of the diameter of rings produced by the adjusted machinery. What are the appropriate null and alternative hypotheses?
Answer: $H_0: \sigma^2 \le 0.03$ and $H_a: \sigma^2 > 0.03$ The engineer's suspicion is that the variance has increased, so the research hypothesis is $\sigma^2 > 0.03$ . This becomes the alternative hypothesis ( $H_a$ ). The null hypothesis ( $H_0$ ) represents the condition that the specification is being met or exceeded in terms of precision (i.e., variance is not larger than specified) and includes the equality case. Therefore, $H_0: \sigma^2 \le 0.03$ .
A researcher is studying a new cognitive therapy. To test its efficacy, the researcher randomly assigns patients with a specific condition to one of two groups: one receiving the new therapy and a control group receiving the standard therapy. The researcher wants to determine if the new therapy has any effect on a standardized measure of patient wellness compared to the standard therapy. Let $\mu_{new}$ and $\mu_{std}$ be the population mean wellness scores for the two groups. What are the null and alternative hypotheses?
Answer: $H_0: \mu_{new} - \mu_{std} = 0$ and $H_a: \mu_{new} - \mu_{std} \neq 0$ The researcher wants to know if the therapy has any effect, which could be positive or negative. This implies a non-directional, or two-tailed, test. The null hypothesis ( $H_0$ ) is that there is no effect, meaning the mean wellness scores for the two groups are the same ( $\mu_{new} = \mu_{std}$ , or $\mu_{new} - \mu_{std} = 0$ ). The alternative hypothesis ( $H_a$ ) is that there is an effect, meaning the means are different ( $\mu_{new} \neq \mu_{std}$ , or $\mu_{new} - \mu_{std} \neq 0$ ).
A website manager knows that, on average, users used to spend 180 seconds on the homepage. After a major redesign, the manager wants to determine if the average time spent on the homepage has decreased. A random sample of user sessions is collected after the redesign. Let $\mu$ be the true mean time users spend on the redesigned homepage. Which hypotheses should the manager test?
Answer: $H_0: \mu \ge 180$ and $H_a: \mu < 180$ When setting up hypothesis tests, you need to carefully translate the research question into statistical language. The manager wants to determine if the average time has decreased from the original 180 seconds, which means they're looking for evidence that the new mean is less than 180. The correct approach is option D: $H_0: \mu \ge 180$ and $H_a: \mu < 180$ . The null hypothesis represents the status quo or "no change" scenario - that the redesign hasn't decreased time (users still spend at least 180 seconds). The alternative hypothesis captures what the manager hopes to prove - that the mean time is now less than 180 seconds. This is a one-tailed test because we're only interested in whether time decreased, not whether it changed in either direction. Option A incorrectly uses the sample mean ( $\bar{x}$ ) instead of the population parameter ( $\mu$ ) in the hypotheses. Hypotheses are always statements about population parameters, not sample statistics. Option B sets up a two-tailed test ( $\mu \neq 180$ ), which would be appropriate if the manager wanted to detect any change in either direction. However, the question specifically asks about a decrease. Option C has the hypotheses backwards. It would test whether time increased after the redesign, which is the opposite of what the manager wants to investigate. Study tip: Always match the alternative hypothesis to the research question's direction. "Decreased," "less than," or "reduced" calls for $H_a: \mu < \text{value}$ , while "increased" needs $H_a: \mu > \text{value}$ , and "changed" or "different" requires $H_a: \mu \neq \text{value}$ .
An agricultural scientist is comparing the mean yield of three different varieties of corn (A, B, and C). The scientist will conduct an Analysis of Variance (ANOVA) to determine if there is a difference in the true mean yields among the three varieties. Which of the following correctly states the null hypothesis ( $H_0$ ) for this ANOVA test?
Answer: $H_0: \mu_A = \mu_B = \mu_C$ In ANOVA, the null hypothesis posits that there is no difference among the group means. For three groups with means $\mu_A$ , $\mu_B$ , and $\mu_C$ , the null hypothesis is that all these population means are equal: $H_0: \mu_A = \mu_B = \mu_C$ . The alternative hypothesis (not asked for here) is that at least one mean is different from the others.
A federal agency sets a guideline that the mean concentration of a certain chemical in industrial wastewater should not exceed 50 parts per million (ppm). The agency conducts regular inspections to find evidence of companies that are in violation of this guideline. When the agency inspects a particular plant, what null and alternative hypotheses should its investigators use?
Answer: $H_0: \mu \le 50$ and $H_a: \mu > 50$ The agency is looking for evidence of a violation. A violation occurs if the mean concentration $\mu$ is greater than the 50 ppm limit. This is the condition the agency seeks to prove, so it becomes the alternative hypothesis: $H_a: \mu > 50$ . The null hypothesis represents the state of compliance with the guideline, which is that the mean concentration is less than or equal to the limit: $H_0: \mu \le 50$ . The burden of proof is on the agency to show that the plant is out of compliance.
A researcher is studying the correlation between the number of hours a student studies per week and their grade point average (GPA). The researcher wants to determine if there is any statistically significant linear relationship between these two variables, but has no prior belief about whether the relationship is positive or negative. Let $\rho$ be the population correlation coefficient. Which are the appropriate hypotheses?
Answer: $H_0: \rho = 0$ and $H_a: \rho \neq 0$ The researcher is testing for any linear relationship, without specifying a direction. This calls for a two-tailed test. The parameter for population correlation is $\rho$ . The null hypothesis for a correlation test is that there is no linear relationship, which corresponds to $\rho = 0$ . The alternative hypothesis is that there is some linear relationship, meaning $\rho \neq 0$ .
A city claims that the proportion of its residents who use public transportation daily is 35%. A sociologist believes this figure is inaccurate and decides to test the claim by surveying a random sample of residents. Which statement correctly formulates the sociologist's alternative hypothesis ( $H_a$ )?
Answer: $H_a$ : The true proportion of city residents who use public transportation is not equal to 0.35. The sociologist believes the city's claim is inaccurate, which suggests a difference in either direction (could be higher or lower). Therefore, the test is two-tailed. The alternative hypothesis reflects the researcher's belief and is a statement about the true population proportion (often denoted $p$ ), not the sample proportion ( $\hat{p}$ ). The correct verbal formulation is that the true proportion is not equal to the claimed value.
To gain approval for a generic drug, a pharmaceutical company must provide evidence that its effect is equivalent to the brand-name drug. Equivalence is defined as the true mean difference in a key therapeutic metric ( $\mu_{generic} - \mu_{brand}$ ) being within a small, clinically insignificant margin, $\Delta$ . The null hypothesis for an equivalence test is that the drugs are meaningfully different. Which pair of hypotheses correctly frames the test for equivalence?
Answer: $H_0: |\mu_{generic} - \mu_{brand}| \ge \Delta$ and $H_a: |\mu_{generic} - \mu_{brand}| < \Delta$ Equivalence testing reverses the standard hypothesis setup. The goal is to find evidence for equivalence, which means the alternative hypothesis should represent the state of equivalence. Equivalence is defined as the absolute difference in means being within a margin $\Delta$ , so $H_a: |\mu_{generic} - \mu_{brand}| < \Delta$ . The null hypothesis represents the state of non-equivalence, meaning the absolute difference is greater than or equal to that margin, $H_0: |\mu_{generic} - \mu_{brand}| \ge \Delta$ . Rejecting the null provides evidence for equivalence.
A community activist claims that the median household income in a certain area is $60,000. An economist suspects this claim is too low. The economist knows that the income distribution in this area is heavily right-skewed. To test this suspicion, which set of hypotheses regarding a population parameter is most appropriate?
Answer: $H_0: M \le 60,000$ and $H_a: M > 60,000$ Because the income distribution is heavily skewed, the median (M) is a more robust measure of central tendency than the mean ( $\mu$ ). Therefore, the hypotheses should be about the population median. The economist suspects the claim of $60,000 is too low, meaning the research hypothesis is that the median is greater than $60,000. This becomes the alternative hypothesis: $H_a: M > 60,000$ . The null hypothesis is that the claim is correct or an overstatement: $H_0: M \le 60,000$ .
Historically, the proportion of students passing a standardized exam at a certain university was 0.75. After the university implemented a new suite of academic support services, an administrator wants to determine if the passing proportion has changed. Let $p$ be the true proportion of students who pass the exam after the implementation of the new services. Which set of hypotheses is most appropriate for the administrator's investigation?
Answer: $H_0: p = 0.75$ and $H_a: p \neq 0.75$ The administrator wants to know if the proportion has changed, which does not imply a specific direction (increase or decrease). This indicates a two-tailed test. The hypotheses must be about the population parameter $p$ , not the sample statistic $\hat{p}$ . The null hypothesis states that there is no change from the historical value, $H_0: p = 0.75$ . The alternative hypothesis states that there is a change, $H_a: p \neq 0.75$ .
A pharmaceutical company has developed a new drug to reduce recovery time from a certain illness. The standard drug has a mean recovery time of 8 days. The company will only market the new drug if it can demonstrate with strong evidence that its new drug's mean recovery time is shorter than the standard drug. Let $\mu_{new}$ represent the true mean recovery time for the new drug. Which of the following are the correct null and alternative hypotheses for the company's clinical trial?
Answer: $H_0: \mu_{new} \ge 8$ and $H_a: \mu_{new} < 8$ The company wants to find evidence that the new drug's recovery time is shorter than the standard 8 days. This is the research claim and becomes the alternative hypothesis ( $H_a: \mu_{new} < 8$ ). The null hypothesis ( $H_0$ ) represents the status quo or the situation of no improvement, and it must contain the condition of equality. Therefore, the null hypothesis is that the mean recovery time is greater than or equal to 8 days ( $H_0: \mu_{new} \ge 8$ ).
A ballot initiative requires a 'supermajority' of 60% of votes to pass. A political polling organization wants to conduct a survey to determine if there is convincing evidence that the initiative will pass. Let $p$ be the true proportion of all voters who support the initiative. Which are the appropriate hypotheses for this poll?
Answer: $H_0: p \le 0.60$ and $H_a: p > 0.60$ The question is about the proportion of voters, so the parameter of interest is $p$ , not $\mu$ . The polling organization is looking for evidence that the initiative will pass, which means the proportion of support is greater than 60% ( $p > 0.60$ ). This is the research question, so it becomes the alternative hypothesis ( $H_a$ ). The null hypothesis represents the opposite case (the initiative will not pass or will exactly meet the threshold), and it includes the equality: $H_0: p \le 0.60$ .
A researcher sets up the hypotheses $H_0: \mu = 50$ and $H_a: \mu > 50$ . After collecting data from a sample of 40 individuals, the researcher calculates a sample mean of $\bar{x} = 53$ . Which of the following statements is a correct interpretation of the role of the hypotheses?
Answer: The hypotheses are a test of whether the true population mean $\mu$ is greater than the hypothesized value of 50. When you encounter hypothesis testing questions, remember that hypotheses are always statements about population parameters, not sample statistics. The entire framework is designed to make inferences about unknown population values based on sample evidence. The correct answer is D because hypothesis tests are fundamentally about determining whether there's sufficient evidence to conclude that a population parameter (like $\mu$ ) differs from a hypothesized value. Here, you're testing whether the true population mean is greater than 50, using the sample mean of 53 as evidence. Let's examine why the other options miss the mark: Option A incorrectly focuses on testing the sample mean itself. The sample mean $\bar{x} = 53$ is an observed fact from your data—there's nothing to "test" about it. You use this sample statistic to make inferences about the unknown population mean. Option B reflects a fundamental misunderstanding of hypothesis testing. You establish your hypotheses before collecting data, not after. The hypotheses represent competing claims about the population parameter, and changing them based on your sample results would invalidate the entire testing procedure. Option C confuses the null hypothesis's meaning. $H_0: \mu = 50$ claims the population mean equals 50, not that your sample mean will be exactly 50. In fact, you expect some sampling variability—your sample mean will likely differ from 50 even if the null hypothesis is true. Key takeaway: Always remember that hypothesis tests make claims about population parameters ( $\mu$ , $p$ , etc.), not sample statistics ( $\bar{x}$ , $\hat{p}$ , etc.). The sample data serves as evidence for or against these population-level claims.

Start this quiz