Null & Alternative Hypotheses - Biostatistics

A researcher is conducting an equivalence trial to show that a generic drug has bioavailability equivalent to the brand name drug. Equivalence is defined as the ratio of mean AUCs being within 80% to 125% (i.e., 0.8 ≤ θ ≤ 1.25 where θ = μ_generic/μ_brand). Which null hypothesis correctly represents the condition that must be rejected to demonstrate equivalence?
Answer: $H_0: \theta < 0.8 \text{ or } \theta > 1.25$ vs $H_a: 0.8 \leq \theta \leq 1.25$ In equivalence testing, the null hypothesis represents non-equivalence (what we want to reject), and the alternative represents equivalence (what we want to demonstrate). The drugs are equivalent if 0.8 ≤ θ ≤ 1.25, so they are non-equivalent if θ < 0.8 or θ > 1.25. Option C correctly places non-equivalence in the null. Option A tests for exact equality, not equivalence within a range. Option B reverses the null and alternative. Option D uses an absolute value formulation that doesn't match the asymmetric equivalence bounds (80%-125% is not symmetric around 1).
A pharmaceutical company claims their new antihypertensive drug reduces systolic blood pressure by at least 15 mmHg compared to placebo. In designing a clinical trial to evaluate this claim, which pair of hypotheses would be most appropriate for regulatory approval purposes?
Answer: $H_0: \mu_{drug} - \mu_{placebo} \geq 15$ vs $H_a: \mu_{drug} - \mu_{placebo} < 15$ For regulatory approval, the burden of proof should be on the company to demonstrate their claim. The null hypothesis should represent the skeptical position (drug does NOT achieve the claimed reduction), while the alternative represents what the company must prove. Option A correctly places the burden on the company to reject the null that the reduction is less than 15 mmHg. Option B tests equality rather than the minimum claim. Option C incorrectly makes it easy for the company by assuming the drug works unless proven otherwise. Option D tests against no effect rather than the specific 15 mmHg claim.
A pharmaceutical company claims that their new antihypertensive medication reduces systolic blood pressure by more than 10 mmHg on average. In designing a clinical trial to test this claim, what should be the null and alternative hypotheses if μ represents the mean reduction in systolic blood pressure?
Answer: $H_0: μ ≤ 10$ ; $H_1: μ > 10$ When you encounter hypothesis testing problems involving company claims or manufacturer assertions, remember that the null hypothesis always represents the status quo or the opposite of what you're trying to prove. The company wants to demonstrate their drug is better than a 10 mmHg reduction, so you're testing whether they can prove superiority. The correct approach is option A: $H_0: μ ≤ 10$ and $H_1: μ > 10$ . The null hypothesis assumes the drug is not superior (reduces blood pressure by 10 mmHg or less), while the alternative hypothesis represents what the company wants to prove (reduction greater than 10 mmHg). This is a one-tailed test because we're specifically testing for superiority in one direction. Option B ( $H_0: μ = 10$ ; $H_1: μ ≠ 10$ ) sets up a two-tailed test that would detect any difference from exactly 10 mmHg, but doesn't address the specific claim of superiority. Option C ( $H_0: μ ≥ 10$ ; $H_1: μ < 10$ ) reverses the logic entirely—it would test whether the drug performs worse than claimed. Option D ( $H_0: μ > 10$ ; $H_1: μ ≤ 10$ ) incorrectly places the claim in the null hypothesis, which violates the fundamental principle that the null represents what you assume until proven otherwise. Study tip: In clinical trials testing superiority claims, always put the "no effect" or "not superior" condition in $H_0$ and the desired outcome in $H_1$ . The burden of proof lies with whoever makes the claim.
A medical device company claims their new glucose meter has measurement bias within ±2 mg/dL of reference values across the physiological range. For regulatory submission, they must demonstrate this claim is true. If μ represents the mean bias, which hypothesis formulation places the appropriate burden of proof on the company?
Answer: $H_0: |\mu| \leq 2$ vs $H_a: |\mu| > 2$ The company must demonstrate their device meets the ±2 mg/dL specification. The burden should be on them to show the bias is acceptable. The null should represent the unacceptable condition (bias exceeds limits), requiring evidence to reject it in favor of acceptability. Option A correctly structures this so rejecting the null demonstrates compliance. Option B makes it too easy by assuming the device is bad unless proven good, which isn't how regulatory approval works. Option C is mathematically equivalent to A but uses non-standard notation. Option D tests only for zero bias, not the ±2 mg/dL tolerance.
A researcher is investigating whether a new therapy increases the median survival time for cancer patients. Historical data shows a median survival time of 18 months. The researcher will only recommend the therapy if there's evidence it provides improvement. If M represents the median survival time with the new therapy, what hypotheses should be formulated?
Answer: $H_0: M ≤ 18$ ; $H_1: M > 18$ When you encounter hypothesis testing problems, the key is understanding what the researcher wants to prove and how hypothesis testing is structured. The researcher seeks evidence that the new therapy improves survival time, meaning they want to show it increases the median beyond the historical 18 months. In hypothesis testing, you always put the status quo or "no effect" condition in the null hypothesis ( $H_0$ ), while the alternative hypothesis ( $H_1$ ) contains what you're trying to prove. Since the researcher will only recommend the therapy if there's evidence of improvement, they need to prove $M > 18$ . This means $H_1: M > 18$ (what they want to demonstrate) and $H_0: M ≤ 18$ (the assumption that the therapy doesn't improve survival or makes it worse). Looking at the wrong answers: Choice A ( $H_0: M = 18$ ; $H_1: M ≠ 18$ ) sets up a two-tailed test that would detect any difference from 18 months, but the researcher specifically wants to show improvement, not just any change. Choice C ( $H_0: M ≥ 18$ ; $H_1: M < 18$ ) tests whether the therapy worsens survival, which is the opposite of what's needed. Choice D ( $H_0: M > 18$ ; $H_1: M ≤ 18$ ) incorrectly puts the desired outcome in the null hypothesis. The correct answer is B because it properly structures a one-tailed test where rejecting $H_0$ provides evidence supporting the therapy's effectiveness. Study tip: Remember that $H_1$ always contains what you're trying to prove, while $H_0$ represents the skeptical position or status quo.
In a bioequivalence study, researchers need to demonstrate that a generic drug has the same mean bioavailability as the brand-name drug. Regulatory guidelines require showing that the ratio of mean bioavailabilities is between 0.8 and 1.25. If R represents this ratio (generic/brand), how should the hypotheses be structured for regulatory approval?
Answer: $H_0: R < 0.8$ or $R > 1.25$ ; $H_1: 0.8 ≤ R ≤ 1.25$ Bioequivalence studies use equivalence testing, which flips traditional hypothesis testing logic. Instead of trying to prove a difference exists, you're trying to prove that two treatments are practically equivalent within specified bounds. In equivalence testing, the null hypothesis represents what you want to reject - that the treatments are NOT equivalent (outside acceptable limits). The alternative hypothesis represents what you want to prove - that they ARE equivalent (within acceptable limits). Since regulatory approval requires demonstrating bioequivalence, you need to structure hypotheses so that rejecting $H_0$ leads to approval. Option C correctly sets $H_0: R < 0.8$ or $R > 1.25$ (non-equivalence) and $H_1: 0.8 ≤ R ≤ 1.25$ (equivalence). By rejecting this null hypothesis, you demonstrate the ratio falls within the acceptable range, supporting regulatory approval. Option A uses traditional two-sided testing, which only tests if the ratio differs from 1, not whether it's within the regulatory range of 0.8-1.25. Option B reverses the logic - it makes equivalence the null hypothesis, which would require proving non-equivalence for rejection rather than proving equivalence. Option D uses incorrect inequality symbols and doesn't properly represent the equivalence bounds as a continuous range. Remember: In equivalence testing, always put what you want to disprove (non-equivalence) as $H_0$ and what you want to prove (equivalence within bounds) as $H_1$ . This ensures that statistical significance leads to the desired regulatory conclusion.
In a multi-center clinical trial, researchers want to test whether the treatment effect (measured as mean reduction in symptoms) is consistent across all centers versus having significant variation between centers. If $σ²_c$ represents the variance in treatment effects between centers, what hypotheses should be tested?
Answer: $H_0: σ²_c = 0$ ; $H_1: σ²_c > 0$ When you encounter questions about variance components in multi-center trials, you're dealing with hypothesis testing for variance parameters. The key insight is understanding what each hypothesis represents and recognizing the directional nature of variance testing. In this scenario, researchers want to determine if treatment effects vary significantly between centers. If $σ²_c = 0$ , this means there's no variation between centers—the treatment effect is consistent everywhere. If $σ²_c > 0$ , there's meaningful variation between centers in how well the treatment works. The correct approach is C: $H_0: σ²_c = 0$ versus $H_1: σ²_c > 0$ . The null hypothesis assumes no between-center variation (homogeneity), while the alternative suggests significant heterogeneity exists. Since variance cannot be negative, we use a one-sided alternative hypothesis testing whether variance is greater than zero. A is incorrect because it uses a two-sided alternative ( $σ²_c ≠ 0$ ), but variance components are always non-negative, making the "less than zero" portion meaningless. B reverses the logic entirely—you never assume variance exists under the null hypothesis when testing for its presence. D includes $σ²_c ≤ 0$ in the null hypothesis, which is nonsensical since variance cannot be negative. Study tip: For variance component testing, always remember that variance is non-negative, so your null hypothesis should assume no variation ( $σ² = 0$ ) and your alternative should test for positive variation ( $σ² > 0$ ). This creates a one-sided test in the positive direction.
A researcher wants to test whether the median time to recovery differs between two treatment groups. Group A receives standard care and Group B receives an experimental treatment. The researcher expects the experimental treatment might reduce recovery time but wants to test for any difference. If $M_A$ and $M_B$ represent the median recovery times, what hypotheses should be tested?
Answer: $H_0: M_A = M_B$ ; $H_1: M_A ≠ M_B$ When testing for differences between groups in biostatistics, the key decision is whether you're looking for any difference (two-tailed test) or a specific directional difference (one-tailed test). The question states the researcher "wants to test for any difference," which signals a two-tailed approach, even though there's an expectation about the direction. The correct hypotheses are those in option A: $H_0: M_A = M_B$ and $H_1: M_A ≠ M_B$ . The null hypothesis assumes no difference between median recovery times, while the alternative hypothesis tests for any difference in either direction. This two-tailed test is appropriate because the researcher wants to detect whether the experimental treatment performs differently from standard care, regardless of whether it's better or worse. Option B ( $H_0: M_A ≤ M_B$ ; $H_1: M_A > M_B$ ) incorrectly assumes we're only testing whether standard care is worse than experimental treatment. Option C ( $H_0: M_A ≥ M_B$ ; $H_1: M_A < M_B$ ) represents a one-tailed test in the expected direction, but this ignores the researcher's stated goal of testing for "any difference." Option D expresses the same hypotheses as A but in difference form rather than equality form—while mathematically equivalent, the equality format in A is the standard convention for stating hypotheses. Remember: when a question mentions testing for "any difference" or "whether groups differ," choose two-tailed hypotheses with equality in the null and inequality (≠) in the alternative, regardless of directional expectations mentioned.
In a clinical trial, researchers want to test whether a new treatment reduces the proportion of patients who experience treatment failure. Historical data shows a failure rate of 0.30. The treatment will be considered effective only if the failure rate drops below 0.25. If p represents the failure rate with the new treatment, what hypotheses should be tested to determine treatment effectiveness?
Answer: $H_0: p ≥ 0.25$ ; $H_1: p < 0.25$ When designing hypothesis tests for clinical effectiveness, you need to carefully consider what question you're actually trying to answer. The key insight here is that the researchers aren't just testing whether the new treatment is different from historical data—they want to prove the treatment meets a specific effectiveness threshold. The correct approach is option B: $H_0: p ≥ 0.25$ ; $H_1: p < 0.25$ . This directly tests whether the failure rate drops below the 0.25 threshold that defines treatment effectiveness. The null hypothesis assumes the treatment is NOT effective (failure rate remains at or above 0.25), while the alternative hypothesis represents what the researchers hope to prove—that the treatment is effective with a failure rate below 0.25. Option A tests whether the new treatment differs from historical rates (0.30) but doesn't address the effectiveness threshold of 0.25. Even if you rejected this null hypothesis, you wouldn't know if the treatment meets the effectiveness criteria. Option C reverses the hypotheses, making it a test to prove the treatment is ineffective—the opposite of what researchers want to demonstrate. Option D only tests against the exact threshold value (0.25) rather than testing whether the rate falls below it, which doesn't properly frame the effectiveness question. Study tip: In clinical effectiveness studies, always identify the threshold that defines "success" and set up your hypotheses to test whether that threshold is met. The alternative hypothesis should reflect what you're trying to prove, and the null should represent the "no effect" or "not effective" scenario.
In a dose-response study, researchers want to test whether increasing the dose of a medication leads to a positive linear relationship between dose (x) and therapeutic effect (y). They specifically want to test if the slope of the regression line is positive. If β represents the population slope coefficient, what hypotheses should be tested?
Answer: $H_0: β ≤ 0$ ; $H_1: β > 0$ When you encounter hypothesis testing in regression analysis, you need to carefully match your research question to the appropriate directional test. This study specifically wants to test whether there's a positive linear relationship, meaning they have a directional hypothesis about the slope. The correct approach is option B: $H_0: β ≤ 0$ ; $H_1: β > 0$ . This is a one-tailed test where the null hypothesis states that the slope is zero or negative (no positive relationship), while the alternative hypothesis states what the researchers want to prove - that the slope is positive. This setup allows you to conclude that increasing dose leads to increased therapeutic effect if you reject the null. Option A ( $H_0: β = 0$ ; $H_1: β ≠ 0$ ) is a two-tailed test that only tests whether there's any linear relationship, not specifically a positive one. You could reject this null even with a negative slope, which doesn't answer the research question. Option C ( $H_0: β ≥ 0$ ; $H_1: β < 0$ ) tests for a negative relationship, which is the opposite of what researchers want to demonstrate. Option D ( $H_0: β < 0$ ; $H_1: β ≥ 0$ ) has the null hypothesis claiming the slope is negative, which is unusual since we typically put the status quo or "no effect" condition in the null hypothesis. Study tip: In regression hypothesis testing, always put what you want to prove in the alternative hypothesis. For directional research questions like "does X increase Y?", use one-tailed tests with the desired direction in $H_1$ .
A quality control analyst must verify that a manufacturing process maintains a defect rate below 2%. The process will be shut down for costly recalibration only if there is strong statistical evidence that the defect rate exceeds this threshold. Production continues unless proven otherwise. Which hypothesis formulation reflects this decision framework?
Answer: $H_0: p \leq 0.02$ vs $H_a: p > 0.02$ When you encounter hypothesis testing problems involving decision frameworks, focus on identifying the "status quo" condition that continues unless there's strong evidence to change it. The key insight is that the null hypothesis always represents the assumption we maintain until proven otherwise. In this scenario, production continues (the status quo) unless there's compelling evidence the defect rate exceeds 2%. This means we assume the process is acceptable ( $p \leq 0.02$ ) until we have strong statistical evidence it's unacceptable ( $p > 0.02$ ). The costly shutdown only happens when we can definitively prove the defect rate is too high. Answer D correctly captures this: $H_0: p \leq 0.02$ vs $H_a: p > 0.02$ . The null hypothesis represents "process is acceptable, keep running," while the alternative represents "process is unacceptable, shut down for recalibration." Answer A uses a two-sided test ( $p \neq 0.02$ ), but we only care if the defect rate is too high, not too low. Answer B reverses the logic entirely—it assumes the process is bad ( $p > 0.02$ ) until proven good, which would mean shutting down by default. Answer C also reverses the framework, assuming we should shut down unless we prove the process is good enough. Remember: in hypothesis testing, the null hypothesis always represents the "innocent until proven guilty" position. Identify what continues by default (here, production), then set that as your null hypothesis. The burden of proof lies with whoever wants to change the status quo.
In a study of vaccine efficacy, researchers want to test whether a new vaccine provides at least 70% protection against infection. The vaccine efficacy is calculated as (1 - relative risk). If E represents the vaccine efficacy, what hypotheses should be tested to evaluate whether the vaccine meets the minimum efficacy requirement?
Answer: $H_0: E ≤ 0.70$ ; $H_1: E > 0.70$ When you encounter vaccine efficacy studies, you're dealing with one-tailed hypothesis tests where researchers want to prove a treatment meets or exceeds a minimum standard. The key is identifying what the researchers are trying to demonstrate versus what they assume under the null hypothesis. In this study, researchers want to prove the vaccine provides at least 70% protection. In hypothesis testing, you set up the null hypothesis as the opposite of what you're trying to prove. Since they want to show $E ≥ 0.70$ , the null hypothesis should assume $E ≤ 0.70$ (the vaccine doesn't meet the standard), and the alternative hypothesis should be $E > 0.70$ (the vaccine exceeds the minimum requirement). This makes option C correct. Option A sets up a two-tailed test ( $E ≠ 0.70$ ) when you need a one-tailed test, and tests for exact equality rather than a minimum threshold. Option B reverses the logic—it assumes the vaccine works well under the null hypothesis, which contradicts the fundamental principle that null hypotheses typically represent "no effect" or "doesn't meet standards." Option D also has the hypotheses backwards and uses incorrect inequality symbols. Remember this pattern: when a study aims to prove something "meets or exceeds" a standard, set your null hypothesis as "fails to meet the standard" (≤) and your alternative as "exceeds the standard" (>). The researchers will try to reject the null to prove their treatment works.
A quality control manager wants to ensure that the mean weight of medication tablets does not deviate from the target weight of 250 mg by more than 5 mg in either direction. The production process will be considered acceptable if the mean weight is between 245 mg and 255 mg. What hypotheses should be tested?
Answer: $H_0: |μ - 250| ≤ 5$ ; $H_1: |μ - 250| > 5$ When you encounter quality control problems with acceptable ranges, you're dealing with equivalence testing rather than traditional hypothesis testing. The key insight is understanding what the null and alternative hypotheses represent in terms of the manufacturing decision. The correct approach is Answer D: $H_0: |μ - 250| ≤ 5$ ; $H_1: |μ - 250| > 5$ . This formulation directly captures the quality control scenario. The null hypothesis states that the mean weight is within the acceptable range (within 5 mg of the target), while the alternative hypothesis represents the undesirable situation where the process has deviated too far from target. The absolute value notation $|μ - 250| ≤ 5$ is mathematically equivalent to $245 ≤ μ ≤ 255$ , but emphasizes the deviation concept. Answer A represents a traditional two-tailed test for whether the mean equals exactly 250 mg, but this doesn't address the acceptable range concept that's central to quality control. Answer B has the logic backwards - it treats being within specifications as the null hypothesis, which would make statistical sense but doesn't align with quality control decision-making where we want to detect deviations from acceptable performance. Answer C also has backwards logic, treating the undesirable state (out of specification) as the null hypothesis. Study tip: In quality control contexts, the null hypothesis typically represents the acceptable or desired state of the process, while the alternative represents the condition requiring corrective action. Look for formulations that test whether parameters fall outside acceptable limits.
A biostatistician is analyzing data from a crossover trial comparing two treatments for chronic pain. The primary endpoint is the difference in pain scores (Treatment B - Treatment A) within each patient. The research question asks whether Treatment B provides superior pain relief. Given that negative values indicate B is better than A, which hypotheses are correctly formulated?
Answer: $H_0: \mu_d \geq 0$ vs $H_a: \mu_d < 0$ The question asks whether Treatment B provides superior pain relief, and negative differences indicate B is better than A. The null should represent no superiority or inferiority of B (μd ≥ 0), while the alternative represents B being superior (μd < 0). Option B correctly captures this directional hypothesis. Option A is two-sided when the research question is specifically about superiority. Option C has the direction backwards (would test if B is worse). Option D has an illogical null that excludes the equality case.
In a pharmacokinetic study, researchers want to establish that a new formulation of a drug is bioequivalent to the reference formulation. Regulatory guidelines require that the ratio of geometric means for AUC (area under the curve) be between 0.80 and 1.25. If R represents the ratio of geometric means (test/reference), what hypotheses should be tested for bioequivalence?
Answer: $H_0: R < 0.80$ or $R > 1.25$ ; $H_1: 0.80 ≤ R ≤ 1.25$ When you encounter bioequivalence testing questions, remember that this is fundamentally different from traditional hypothesis testing. Instead of trying to prove a difference exists, you're trying to prove that two formulations are equivalent within acceptable regulatory limits. In bioequivalence studies, you want to demonstrate that the test formulation performs similarly to the reference formulation. The regulatory requirement that the ratio R must fall between 0.80 and 1.25 defines the acceptable equivalence range. To prove bioequivalence, you must show that R falls within this range. The correct approach is option C: $H_0: R < 0.80$ or $R > 1.25$ ; $H_1: 0.80 ≤ R ≤ 1.25$ . This sets up the null hypothesis as "the formulations are NOT equivalent" (ratio falls outside acceptable limits) and the alternative hypothesis as "the formulations ARE equivalent" (ratio falls within acceptable limits). When you reject this null hypothesis, you conclude bioequivalence. Option A represents traditional superiority testing, which tests for any difference rather than equivalence. Option B incorrectly makes bioequivalence the null hypothesis—you can't "prove" a null hypothesis, only fail to reject it. Option D uses the wrong inequality symbols and doesn't properly represent the bioequivalence range. Remember this key principle: in bioequivalence testing, you must flip the traditional hypothesis structure. Set up your null hypothesis as "non-equivalence" so that rejecting it allows you to conclude equivalence with statistical confidence. This approach protects against falsely claiming bioequivalence when insufficient evidence exists.
A researcher wants to test whether a new screening method has a specificity greater than the current standard of 0.85. However, implementing the new method is costly and will only be adopted if there is strong evidence of superiority. What hypotheses should be formulated to reflect the appropriate burden of proof?
Answer: $H_0: p_{new} \leq 0.85$ vs $H_a: p_{new} > 0.85$ Since the new method will only be adopted with strong evidence of superiority, the burden of proof should be on demonstrating improvement. The null should represent the conservative position (new method is not better), requiring evidence to reject it in favor of superiority. Option B correctly formulates this with the null including equality and inferiority. Option A only tests equality, not the full range of non-superiority. Option C reverses the burden inappropriately. Option D doesn't incorporate the specific threshold of 0.85 that defines adequate performance.
A vaccine trial is designed to test whether the new vaccine has efficacy of at least 70%. Vaccine efficacy is defined as VE = 1 - (attack rate in vaccinated)/(attack rate in unvaccinated). Regulators will only approve the vaccine if there is convincing evidence it meets this threshold. How should the hypotheses be formulated in terms of the relative risk (RR = attack rate ratio)?
Answer: $H_0: RR \geq 0.30$ vs $H_a: RR < 0.30$ When you encounter vaccine efficacy problems in biostatistics, you need to carefully translate between vaccine efficacy (VE) and relative risk (RR), then properly set up hypotheses based on the regulatory burden of proof. The key relationship is: $VE = 1 - RR$ , where $RR = \frac{\text{attack rate in vaccinated}}{\text{attack rate in unvaccinated}}$ . If the vaccine has 70% efficacy, then $0.70 = 1 - RR$ , so $RR = 0.30$ . Since regulators will only approve the vaccine with "convincing evidence" of at least 70% efficacy, the burden is on the vaccine to prove itself. This means we assume the vaccine is ineffective (doesn't meet the 70% threshold) until proven otherwise. Therefore, the null hypothesis should state that the vaccine is inadequate: $H_0: VE \leq 0.70$ or equivalently $H_0: RR \geq 0.30$ . We reject this only with strong evidence that $RR < 0.30$ (meaning $VE > 0.70$ ). Answer C correctly states $H_0: RR \geq 0.30$ vs $H_a: RR < 0.30$ . Answer A uses the wrong direction for the burden of proof—it assumes the vaccine works until proven otherwise. Answer B flips the inequality signs, making it easier to approve an ineffective vaccine. Answer D uses a two-sided test when we specifically need one-sided evidence of superiority over the 70% threshold. Remember: in regulatory settings, the intervention must prove itself against a skeptical null hypothesis. The burden of proof determines which direction your inequalities should face.
A researcher wants to test whether the variance in blood glucose measurements from a new glucose meter is significantly different from the variance of the standard meter, which is 25 mg²/dL². The new meter should be adopted only if it shows improved precision (lower variance). If $σ²$ represents the variance of the new meter, what hypotheses should be tested?
Answer: $H_0: σ² ≥ 25$ ; $H_1: σ² < 25$ When you encounter hypothesis testing questions about variance, you need to carefully distinguish between two-tailed tests (testing for any difference) and one-tailed tests (testing for a specific direction of difference). The key here is understanding what the researcher actually wants to demonstrate. The correct answer is B because the researcher has a specific goal: to adopt the new meter only if it shows "improved precision," meaning lower variance. This creates a directional research question. In hypothesis testing, you set up the null hypothesis as the opposite of what you're trying to prove. Since the researcher wants to demonstrate that the new meter has lower variance ( $σ² < 25$ ), the null hypothesis should state that the variance is greater than or equal to 25 ( $H_0: σ² ≥ 25$ ), with the alternative being $H_1: σ² < 25$ . Answer A represents a two-tailed test, which would be appropriate if you simply wanted to detect any difference in variance, not specifically an improvement. This doesn't match the researcher's stated goal of only adopting the meter if it's better. Answer C has the hypotheses backwards - it would test whether the new meter is worse (higher variance), which contradicts the research objective. Answer D incorrectly places the inequality in the null hypothesis, violating the standard convention that null hypotheses typically contain equality. Study tip: Remember that one-tailed tests are used when you have a specific directional prediction. Always set your null hypothesis as the opposite of what you're trying to prove, and include equality in the null hypothesis.
A clinical laboratory is validating a new assay and needs to demonstrate that the coefficient of variation (CV) is no more than 5% to meet regulatory standards. The validation will be expensive to repeat if the assay fails to meet this requirement. How should the hypotheses be structured?
Answer: $H_0: CV \leq 0.05$ vs $H_a: CV > 0.05$ The laboratory needs to demonstrate compliance with the standard (CV ≤ 5%). The burden should be on proving the assay is acceptable. The null should represent the unacceptable condition (CV too high), requiring evidence to reject it in favor of acceptability. Option B correctly structures this so that rejecting the null provides evidence the assay meets the standard. Option A tests only equality. Option C makes it too easy to 'prove' compliance by assuming the assay is good unless proven bad. Option D has an illogical null hypothesis that excludes the boundary value.
In a study comparing mean cholesterol levels between two populations, Population A (urban) and Population B (rural), researchers hypothesize that urban living leads to higher cholesterol levels. If $μ_A$ and $μ_B$ represent the population means, what hypotheses should be tested?
Answer: $H_0: μ_A - μ_B ≤ 0$ ; $H_1: μ_A - μ_B > 0$ When you encounter hypothesis testing questions, the key is translating the research question into proper null and alternative hypotheses. The research hypothesis here states that "urban living leads to higher cholesterol levels," which means researchers expect $μ_A > μ_B$ . In hypothesis testing, the null hypothesis ( $H_0$ ) represents the status quo or "no effect" position, while the alternative hypothesis ( $H_1$ ) represents what the researchers are trying to prove. Since researchers hypothesize that urban populations have higher cholesterol, they want to test whether $μ_A > μ_B$ . The null hypothesis must therefore represent the opposite: that urban populations do not have higher cholesterol levels, or $μ_A ≤ μ_B$ . Answer B correctly sets up this one-tailed test: $H_0: μ_A - μ_B ≤ 0$ (urban cholesterol is less than or equal to rural) and $H_1: μ_A - μ_B > 0$ (urban cholesterol is greater than rural). Answer A represents a two-tailed test, which would be appropriate if researchers simply wanted to test whether the populations differ in any direction, not specifically that urban is higher. Answer C uses correct notation but places the directional hypothesis in $H_0$ instead of $H_1$ , which violates hypothesis testing principles. Answer D tests the opposite direction entirely, hypothesizing that rural populations have higher cholesterol. Study tip: Always identify the research direction first, then set up $H_1$ to match that direction and $H_0$ as its complement. The alternative hypothesis should reflect what the researchers are trying to prove, not the null position.

Start this quiz