Biostatistics Quiz: Ci For Differences
20 questions · exam conditions
0:00
Ci For DifferencesQuestion 1 of 20

A researcher compares response rates between two survey methods. Method 1: 340 responses from 800 contacts (42.5%). Method 2: 285 responses from 750 contacts (38.0%). The 95% confidence interval for the difference in response rates (Method 1 - Method 2) is calculated as (0.008, 0.082). If the researcher wanted to test whether Method 1 is superior to Method 2 at α = 0.05, what conclusion should be drawn?

The confidence interval suggests superiority, but a formal hypothesis test is needed
Fail to reject the null hypothesis; insufficient evidence that Method 1 is superior
Reject the null hypothesis; Method 1 is significantly superior with p < 0.05
Accept the alternative hypothesis; the difference of 4.5% is practically significant
← Back to quizzes

Biostatistics Quiz

Biostatistics Quiz: Ci For Differences

Practice Ci For Differences in Biostatistics with focused quiz questions that help you check what you know, review explanations, and build confidence with test-style prompts.

What this quiz covers

This quiz focuses on Ci For Differences, giving you a quick way to practice the rules, question types, and explanations that matter most for Biostatistics.

How to use this quiz

Try each quiz question before looking at the correct answer. Use the explanations to review missed ideas, then come back to similar questions until the pattern feels familiar.

All questions

Question 1

A researcher compares response rates between two survey methods. Method 1: 340 responses from 800 contacts (42.5%). Method 2: 285 responses from 750 contacts (38.0%). The 95% confidence interval for the difference in response rates (Method 1 - Method 2) is calculated as (0.008, 0.082). If the researcher wanted to test whether Method 1 is superior to Method 2 at α = 0.05, what conclusion should be drawn?

  1. The confidence interval suggests superiority, but a formal hypothesis test is needed
  2. Fail to reject the null hypothesis; insufficient evidence that Method 1 is superior
  3. Reject the null hypothesis; Method 1 is significantly superior with p < 0.05 (correct answer)
  4. Accept the alternative hypothesis; the difference of 4.5% is practically significant
Explanation: When you encounter questions about comparing two proportions with confidence intervals, remember that the confidence interval and hypothesis testing are mathematically equivalent - they'll always give you the same conclusion about statistical significance. Here's the key insight: since we're testing whether Method 1 is superior (a one-tailed test), we need to check if the entire confidence interval for the difference lies above zero. The 95% confidence interval (0.008, 0.082) shows that we're 95% confident the true difference in response rates falls between 0.8% and 8.2%. Since zero is not included in this interval, we can reject the null hypothesis that there's no difference between methods. For a one-tailed test at α = 0.05, this translates to p < 0.05, making Method 1 significantly superior. Let's examine why the other options miss the mark: A) suggests we need additional testing, but the confidence interval already provides all the statistical evidence needed for hypothesis testing. B) incorrectly concludes we should fail to reject the null hypothesis, but since zero isn't in our interval, we actually have strong evidence against the null. D) confuses statistical significance with practical significance - while the difference may be statistically significant, whether 4.5% is "practically significant" depends on context, not statistics. Study tip: Remember that if zero falls outside your confidence interval for a difference, you have statistical significance at your chosen alpha level. The confidence interval IS your hypothesis test - you don't need separate procedures.

Question 2

Two hospitals compare their surgical infection rates. Hospital A had 15 infections in 500 surgeries. Hospital B had 8 infections in 300 surgeries. When constructing a 95% confidence interval for the difference in infection rates (A - B), the calculated interval is (-0.005, 0.025). What additional consideration is most important for interpreting this result?

  1. The sample sizes are too small for normal approximation; exact methods should be used instead
  2. The confidence level should be increased to 99% given the clinical importance of infection rates
  3. Hospital characteristics should be considered as potential confounders affecting the comparison validity
  4. The low event rates may violate normality assumptions despite adequate overall sample sizes (correct answer)
Explanation: When comparing proportions between groups, you need to consider both the overall sample size and the actual number of events observed. While the total sample sizes here (500 and 300) seem adequate, the key issue is the low number of infections: only 15 and 8 events respectively. The correct answer is D because when event rates are low (even with reasonable total sample sizes), the sampling distribution of the proportion may not follow a normal distribution well enough for standard confidence interval methods. Infection rates of 3% (15/500) and 2.7% (8/300) represent relatively rare events. The normal approximation for proportions typically requires both np≥5np \geq 5 and n(1−p)≥5n(1-p) \geq 5 for each group. While Hospital A meets this (500×0.03=15500 \times 0.03 = 15), Hospital B is borderline (300×0.027=8.1300 \times 0.027 = 8.1), and the approximation becomes questionable. Option A is incorrect because the total sample sizes are actually reasonable for normal approximation methods in most circumstances. Option B misses the statistical issue entirely—increasing confidence level doesn't address the underlying distributional concerns. Option C, while potentially relevant for causal interpretation, doesn't address the immediate statistical validity of the confidence interval calculation itself. When you encounter proportion comparisons with low event rates, always check whether the normal approximation assumptions are satisfied. Consider exact methods (like Fisher's exact test) or continuity corrections when dealing with rare events, even if overall sample sizes appear adequate.

Question 3

A study compares mean cholesterol levels between vegetarians (n=40, mean=195 mg/dL, SD=25 mg/dL) and non-vegetarians (n=45, mean=218 mg/dL, SD=30 mg/dL). The 95% confidence interval for the difference in means (vegetarian - non-vegetarian) is (-35.2, -10.8). What assumption was most likely violated if this interval is invalid?

  1. The populations have equal variances
  2. The samples are normally distributed
  3. The samples are independent of each other (correct answer)
  4. The population means are different
  5. The sample sizes are sufficiently large
Explanation: When you encounter confidence intervals for comparing two group means, you're dealing with a two-sample t-test framework that relies on several critical assumptions. The validity of your interval depends entirely on whether these assumptions hold. The most fundamental assumption is that your samples must be independent of each other (C). This means the cholesterol measurement from any vegetarian cannot influence or be related to the measurement from any non-vegetarian. If this assumption is violated—say, if some participants were married couples where spouses influence each other's diets, or if measurements were taken from the same individuals at different time points—the standard error calculations become invalid, making your confidence interval meaningless. Let's examine why the other options aren't the primary concern here. Equal variances (A) is an assumption, but modern statistical methods use Welch's t-test that adjusts for unequal variances, so this rarely invalidates results completely. Normal distributions (B) matter, but with sample sizes of 40 and 45, the Central Limit Theorem makes the sampling distribution approximately normal regardless of the underlying population distribution. Different population means (D) isn't an assumption at all—it's actually what we're trying to detect! The key insight is that mathematical adjustments can handle unequal variances and non-normal distributions, but no statistical technique can fix dependence between supposedly independent samples. Study tip: Always check independence first when evaluating statistical validity. Ask yourself: "Could any observation in group 1 be related to any observation in group 2?" If yes, independence is violated.

Question 4

In a clinical trial, the success rate for Drug A was 72% (36 out of 50 patients) and for Drug B was 64% (32 out of 50 patients). If the 95% confidence interval for the difference in success rates (A - B) is (-0.12, 0.28), what can be concluded about the difference between the drugs?

  1. Drug A is significantly better than Drug B at α = 0.05 level
  2. Drug B is significantly better than Drug A at α = 0.05 level
  3. There is no significant difference between drugs at α = 0.05 level (correct answer)
  4. The confidence interval is too wide to make any conclusion
  5. The sample sizes are too small for valid inference
Explanation: When you encounter confidence intervals for differences between two groups, you're testing whether there's a meaningful difference between them. The key insight is that if zero falls within the confidence interval, you cannot conclude there's a significant difference at your chosen alpha level. The 95% confidence interval for the difference in success rates (Drug A - Drug B) is (-0.12, 0.28). Since this interval contains zero, we cannot reject the null hypothesis that there's no difference between the drugs. Even though Drug A appears to perform better (72% vs 64%), this 8 percentage point difference could easily be due to random variation rather than a true treatment effect. Answer A is incorrect because claiming Drug A is significantly better would require the entire confidence interval to be above zero. Since the interval extends down to -0.12, we cannot make this conclusion. Answer B is wrong for the same reason in reverse – the interval extends up to 0.28, so we cannot conclude Drug B is better. Answer D misunderstands the purpose of confidence intervals. The width doesn't prevent us from drawing conclusions; rather, a wide interval that contains zero specifically tells us there's insufficient evidence of a difference. Answer C correctly recognizes that when a 95% confidence interval for a difference contains zero, there's no significant difference at the α = 0.05 level. Remember this rule: if zero is in the confidence interval for a difference, there's no significant difference. This applies whether you're comparing means, proportions, or any other measures between groups.

Question 5

Two hospitals are compared for their infection rates. Hospital A had 18 infections among 300 patients, while Hospital B had 25 infections among 400 patients. When calculating the 95% confidence interval for the difference in infection rates (A - B), which of the following represents the correct point estimate and its interpretation?

  1. -0.0025; Hospital A has a slightly lower infection rate (correct answer)
  2. 0.0025; Hospital A has a slightly higher infection rate
  3. -0.0025; Hospital B has a significantly lower infection rate
  4. 0.0375; Hospital A has a much higher infection rate
  5. -0.0375; Hospital B has a much higher infection rate
Explanation: When comparing two proportions, you need to calculate each hospital's infection rate first, then find their difference. This type of question tests your ability to compute point estimates for comparing population proportions. Let's calculate the infection rates: Hospital A has 18/300 = 0.06 (6% infection rate), while Hospital B has 25/400 = 0.0625 (6.25% infection rate). The point estimate for the difference (A - B) is 0.06 - 0.0625 = -0.0025. This negative value indicates Hospital A has a slightly lower infection rate than Hospital B. Looking at the answer choices: A correctly identifies the point estimate as -0.0025 and properly interprets that Hospital A has a slightly lower rate. B has the wrong sign - a positive 0.0025 would mean Hospital A had a higher rate, which contradicts our calculation. C uses the correct point estimate but incorrectly states that Hospital B has a "significantly" lower rate, when actually Hospital A has the lower rate (and we can't determine statistical significance from just the point estimate). D provides a completely incorrect value of 0.0375, which doesn't match any meaningful calculation from the given data. Study tip: Always calculate both individual rates before finding their difference, and remember that the sign of your difference tells you which group has the higher value. Also, avoid confusing point estimates with statistical significance - you need confidence intervals or hypothesis tests to make claims about significance.

Question 6

In a randomized trial, Group 1 (n=60) had a mean response of 24.5 units (SD=5.2), while Group 2 (n=55) had a mean response of 21.8 units (SD=4.9). The researcher wants to test if there is a difference and constructs a 95% CI for the difference in means. If the calculated standard error is 0.96, but the researcher mistakenly uses z=1.96 instead of the appropriate t-value, how will this affect the confidence interval?

  1. The interval will be too narrow, increasing Type I error risk (correct answer)
  2. The interval will be too wide, increasing Type II error risk
  3. The interval will be too narrow, decreasing power
  4. The interval will be approximately correct due to large sample sizes
  5. The interval will be shifted but maintain correct width
Explanation: When comparing two independent groups with continuous outcomes, you need to decide between using a z-distribution or t-distribution for your confidence interval. This choice depends on whether the population standard deviation is known and the sample sizes involved. In this scenario, you're estimating the population standard deviations from sample data, which means you should use a t-distribution. With degrees of freedom df=n1+n2−2=60+55−2=113df = n_1 + n_2 - 2 = 60 + 55 - 2 = 113, the appropriate t-value would be approximately 1.98, which is larger than the z-value of 1.96. Using the smaller z-value creates a confidence interval that's too narrow. A narrower confidence interval means you're more likely to conclude there's a significant difference when there might not be one, which increases your Type I error risk (falsely rejecting a true null hypothesis). This makes option A correct. Option B is wrong because using z creates a narrower, not wider interval. Option C incorrectly links the narrow interval to decreased power - while a narrow interval does increase Type I error risk, power relates to Type II errors and sample size considerations. Option D suggests the approximation is adequate, but even with reasonably large samples, using z instead of t systematically underestimates the critical value needed. Key takeaway: Always use the t-distribution when estimating population parameters from sample data, regardless of sample size. The difference between t and z values decreases as sample size increases, but using z will always make your confidence intervals slightly too narrow, inflating Type I error rates.

Question 7

Two diet programs are compared for weight loss effectiveness. Program A participants (n=45) lost an average of 8.3 kg (SD=2.1 kg), while Program B participants (n=50) lost an average of 6.7 kg (SD=2.4 kg). When constructing a confidence interval for the difference in mean weight loss, which assumption is most critical to verify given these data?

  1. The weight loss distributions are exactly normal in both groups
  2. The population variances are exactly equal between groups
  3. The participants were randomly assigned to programs (correct answer)
  4. The measurement precision is identical for both groups
  5. The follow-up periods are identical for all participants
Explanation: When comparing means between two groups using confidence intervals or hypothesis tests, you're essentially making inferences about population parameters based on sample data. The validity of these inferences depends critically on how the data were collected and whether certain statistical assumptions are met. The most fundamental requirement is that your sample must be representative of the populations you want to compare. This is achieved through proper study design, particularly random assignment (answer C). Without random assignment, systematic differences between groups could confound your results. For example, if people self-selected their diet program, those choosing Program A might be more motivated or have different baseline characteristics, making it impossible to attribute weight loss differences to the programs themselves. Random assignment ensures that any observed difference is likely due to the intervention, not pre-existing group differences. Let's examine why the other assumptions are less critical: Answer A is incorrect because confidence intervals are robust to non-normality, especially with moderate sample sizes like these (n=45, 50) due to the Central Limit Theorem. Answer B is wrong because unequal variances can be handled with statistical adjustments like Welch's t-test - notice the standard deviations here are reasonably similar anyway (2.1 vs 2.4). Answer D is incorrect because different measurement precision between groups, while potentially problematic, is rarely the most critical assumption and can often be addressed analytically. Study tip: In biostatistics questions about group comparisons, always prioritize study design issues (randomization, bias, confounding) over mathematical assumptions. Proper randomization is the foundation that makes statistical inference meaningful.

Question 8

A researcher studying exercise adherence finds that 42 out of 75 participants in a supervised program completed the full regimen, while 28 out of 65 participants in a self-directed program completed it. When calculating the 95% confidence interval for the difference in completion rates (supervised - self-directed), the researcher gets (-0.013, 0.347). What does the lower bound of this interval indicate?

  1. The supervised program could have a completion rate 1.3% lower than the self-directed program (correct answer)
  2. The self-directed program is definitely superior to the supervised program
  3. The difference in completion rates is not practically significant
  4. There is a 1.3% chance that the supervised program is worse
  5. The margin of error in the study is 1.3%
Explanation: When you encounter confidence intervals for differences between proportions, focus on what each bound tells you about the range of plausible values for the true difference. First, let's calculate the completion rates: supervised program = 42/75 = 0.56 (56%), self-directed program = 28/65 = 0.43 (43%). The confidence interval (-0.013, 0.347) represents plausible values for (supervised rate - self-directed rate). The lower bound of -0.013 means the true difference could be as low as -1.3 percentage points. Since this is negative, it indicates the supervised program could actually have a completion rate 1.3% lower than the self-directed program. This makes choice A correct. Choice B is wrong because the interval includes positive values (up to 34.7%), meaning the supervised program could be superior. A confidence interval spanning zero never proves one treatment is "definitely" better. Choice C misinterprets the interval - a 1.3% difference might seem small, but practical significance depends on context and cost-benefit considerations, not just the numerical value. Choice D confuses the confidence interval with probability - the -0.013 represents a potential difference magnitude, not a probability that one program is worse. Study tip: Remember that confidence intervals for differences tell you the range of plausible values for the true difference. When the interval includes zero, neither treatment is definitively superior. Always interpret the bounds as potential effect sizes, not probabilities.

Question 9

A clinical study compares the mean time to symptom relief between two treatments. Treatment 1: n=28, mean=4.2 hours, SD=1.1 hours. Treatment 2: n=32, mean=5.1 hours, SD=1.3 hours. The researcher calculates a 95% confidence interval assuming equal variances and gets (-1.65, -0.15). If the assumption of equal variances is violated, how would this most likely affect the confidence interval?

  1. The interval would become wider and the degrees of freedom would decrease (correct answer)
  2. The interval would become narrower and more precise
  3. The interval center would shift but the width would remain the same
  4. The interval would become symmetric around zero
  5. The interval would maintain the same width but use different critical values
Explanation: When comparing two means with a t-test, you must decide whether to assume equal variances between groups. This choice affects both your degrees of freedom calculation and the width of your confidence interval. The correct approach here is A. When equal variances are assumed (pooled variance method), you get more degrees of freedom: df=n1+n2−2=28+32−2=58df = n_1 + n_2 - 2 = 28 + 32 - 2 = 58. But if this assumption is violated and you switch to Welch's t-test (unequal variances), the degrees of freedom decrease substantially—often to something much closer to the smaller sample size. Fewer degrees of freedom means a larger critical t-value, which widens the confidence interval. B is wrong because violating the equal variance assumption doesn't make your interval more precise—it makes it less precise due to the decreased degrees of freedom and larger critical value. C is incorrect because while the center might shift slightly due to different calculations, the width definitely changes. The primary effect is interval widening, not just shifting. D makes no sense because the interval was already clearly not symmetric around zero (it's entirely negative: -1.65 to -0.15), and changing the variance assumption wouldn't suddenly make the treatment difference disappear. Study tip: Remember that statistical assumptions aren't just academic—violating them usually makes your results less precise, not more. When you lose an assumption like equal variances, you typically pay a price in terms of wider intervals and reduced power.

Question 10

A pharmaceutical study compares the proportion of patients experiencing nausea as a side effect between Drug A (15 out of 120 patients) and Drug B (28 out of 180 patients). The researcher wants to construct a 99% confidence interval for the difference in proportions (A - B) but is unsure whether to use a normal approximation. Which condition is most important to verify for the validity of the normal approximation?

  1. Each sample has at least 30 total observations
  2. The success-failure condition: np ≥ 10 and n(1-p) ≥ 10 for each group (correct answer)
  3. The difference in sample sizes is less than 50%
  4. The pooled proportion is between 0.1 and 0.9
  5. The total number of successes across both groups exceeds 20
Explanation: When working with confidence intervals for differences in proportions, the validity of the normal approximation depends on having enough successes and failures in each sample to ensure the sampling distribution is approximately normal. The success-failure condition requires that for each group, both np≥10np \geq 10 and n(1−p)≥10n(1-p) \geq 10. Let's check this for our data: Drug A has 15 successes and 105 failures (both ≥ 10), while Drug B has 28 successes and 152 failures (both ≥ 10). This condition ensures the sampling distribution of each proportion is approximately normal, which is essential for constructing valid confidence intervals. Option A is incorrect because the "rule of 30" applies to means, not proportions. For proportions, you need adequate successes and failures regardless of total sample size. Option C is wrong because there's no requirement about relative sample sizes for proportion comparisons—unequal sample sizes are perfectly acceptable. Option D incorrectly focuses on the pooled proportion's value. While extremely small or large pooled proportions can be problematic, the specific range of 0.1 to 0.9 isn't a standard condition, and the pooled proportion is primarily used in hypothesis testing, not confidence interval construction. The correct answer is B because it identifies the fundamental requirement for normal approximation validity in proportion problems. Study tip: For any proportion-based inference, always verify the success-failure condition first. Count actual successes and failures in your data—if any group has fewer than 10 successes or fewer than 10 failures, consider alternative methods like exact binomial procedures.

Question 11

A researcher calculates a 99% confidence interval for the difference in mean test scores between two teaching methods. Method A: n=25, mean=78.4, SD=12.1. Method B: n=30, mean=82.7, SD=10.8. If the confidence interval is (-12.8, 4.2), what would happen to the interval width if the confidence level were changed to 95%?

  1. The interval would become narrower by approximately 35% (correct answer)
  2. The interval would become wider by approximately 25%
  3. The interval would become narrower by approximately 25%
  4. The interval width would remain approximately the same
  5. The interval would become wider by approximately 35%
Explanation: When you encounter confidence interval questions involving changes in confidence levels, remember that the relationship between confidence level and interval width follows a predictable pattern based on critical values from the t-distribution. The current 99% confidence interval (-12.8, 4.2) has a width of 4.2 - (-12.8) = 17.0. To determine how this changes at 95% confidence, you need to compare the critical t-values. With degrees of freedom = (25-1) + (30-1) = 53, the critical value for 99% confidence is approximately 2.67, while for 95% confidence it's approximately 2.01. The ratio is 2.01/2.67 = 0.753, meaning the new interval will be about 75% of the original width. This represents a reduction of approximately 25%, but since we're going from a larger reduction factor, the actual reduction is closer to 35%. Looking at the distractors: Answer B incorrectly suggests the interval becomes wider, which contradicts the fundamental principle that lower confidence levels produce narrower intervals. Answer C calculates the approximate 25% reduction you might expect from a quick mental calculation, but this underestimates the actual change because it doesn't account for the specific t-distribution critical values at these confidence levels. Answer D is wrong because confidence level changes always affect interval width significantly. Study tip: Always remember that confidence level and interval width move in the same direction - higher confidence requires wider intervals to capture the parameter with greater certainty. Practice memorizing common critical values to speed up these calculations.

Question 12

Two hospitals are being compared for patient satisfaction scores on a 100-point scale. Hospital A (n=85, mean=78.3, SD=12.4) and Hospital B (n=90, mean=74.1, SD=14.2). A 95% confidence interval for the difference in mean satisfaction (A - B) is calculated as (-0.8, 9.2). If a difference of at least 5 points is considered clinically meaningful, what conclusion is most appropriate?

  1. Hospital A is significantly better than Hospital B with clinical importance
  2. There is no significant difference, and any difference lacks clinical importance
  3. There is no significant difference, but the difference could be clinically meaningful (correct answer)
  4. Hospital A is significantly better, but the difference is not clinically meaningful
  5. The sample sizes are too small to determine clinical significance
Explanation: When interpreting confidence intervals, you need to distinguish between statistical significance and clinical significance. Statistical significance asks whether there's convincing evidence of any difference, while clinical significance asks whether that difference matters in practice. The 95% confidence interval (-0.8, 9.2) for the difference in satisfaction scores (Hospital A - Hospital B) includes zero. This means we cannot rule out the possibility that there's no difference between hospitals - hence, no statistically significant difference exists. However, the interval also includes values above 5 points (ranging up to 9.2), which would be clinically meaningful according to the problem's criteria. Answer C correctly captures both aspects: no statistical significance (because the interval includes zero) but potential clinical meaningfulness (because the interval includes differences ≥5 points). Answer A is wrong because there's no statistical significance - the confidence interval includes zero, meaning we can't conclude Hospital A is "significantly better." Answer B incorrectly claims any difference lacks clinical importance, but the upper bound of 9.2 points exceeds the 5-point clinical threshold. Answer D contradicts itself by claiming Hospital A is "significantly better" when the confidence interval includes zero, indicating no statistical significance. Remember this key distinction: statistical significance depends on whether the confidence interval excludes zero (or null value), while clinical significance depends on whether the interval includes values that meet your predetermined threshold for meaningful difference. A result can be statistically non-significant yet still suggest potential clinical importance.

Question 13

A medical device company tests two glucose meters for accuracy. Device X shows a mean reading of 105.2 mg/dL (n=30, SD=3.8) and Device Y shows 103.7 mg/dL (n=35, SD=4.1) when measuring the same blood samples. The 95% confidence interval for the difference (X - Y) is calculated as (-1.2, 4.2). What is the most appropriate interpretation?

  1. Device X is significantly more accurate than Device Y
  2. Device Y is significantly more accurate than Device X
  3. There is insufficient evidence of a significant difference between devices (correct answer)
  4. Device X consistently reads higher than Device Y
  5. The confidence interval is too wide to be clinically useful
Explanation: When you encounter confidence intervals for differences between two groups, you're testing whether there's a statistically significant difference between them. The key insight is what the confidence interval tells you about the null hypothesis (no difference between groups). The 95% confidence interval for the difference (X - Y) is (-1.2, 4.2). Since this interval includes zero, we cannot conclude there's a significant difference between the devices. Zero represents the null hypothesis that there's no true difference between Device X and Device Y. When zero falls within our confidence interval, we fail to reject the null hypothesis. Answer C is correct because the confidence interval spanning zero indicates insufficient evidence of a significant difference. We cannot distinguish between the devices' accuracy based on this data. Answer A is wrong because statistical significance requires the entire confidence interval to be above zero (positive difference). Since our interval includes negative values, X is not significantly more accurate than Y. Answer B is incorrect for the same reason in reverse - the interval would need to be entirely below zero to conclude Y is significantly more accurate than X. Answer D is tempting because the sample mean for X (105.2) is higher than Y (103.7), but this misses the statistical inference concept. The confidence interval shows this observed difference could easily be due to random sampling variation rather than a true systematic difference. Study tip: When interpreting confidence intervals for differences, always check if zero is included. If yes, there's no significant difference regardless of which sample mean appears larger.

Question 14

A clinical trial comparing two treatments found a difference in mean recovery time of -3.2 days (Treatment A minus Treatment B) with a standard error of 1.4 days. If the 95% confidence interval for this difference is (-6.04, -0.36), what critical t-value was used in the calculation?

  1. 2.03 (correct answer)
  2. 1.96
  3. 2.57
  4. 1.83
  5. 2.45
Explanation: When you encounter confidence interval problems, you're working with the fundamental relationship between the margin of error, standard error, and critical values. The confidence interval formula is: point estimate ± (critical value × standard error). To find the critical t-value, you need to work backwards from the given information. The point estimate is -3.2 days, and the confidence interval is (-6.04, -0.36). The margin of error equals the distance from the point estimate to either endpoint: ∣−3.2−(−6.04)∣=2.84|-3.2 - (-6.04)| = 2.84 days. Since margin of error = critical value × standard error, you can solve: 2.84=t×1.42.84 = t \times 1.4, giving you t=2.84÷1.4=2.03t = 2.84 ÷ 1.4 = 2.03. This confirms answer A is correct. Looking at the wrong answers: B (1.96) is the critical z-value for 95% confidence with large samples, but this problem uses a t-distribution, likely due to a smaller sample size. C (2.57) would be a t-value for 95% confidence with very few degrees of freedom (around 4-5), suggesting an unrealistically small sample. D (1.83) doesn't correspond to standard confidence levels and would give a margin of error that doesn't match the given interval. Study tip: Always check whether a problem uses z or t distributions. If you see "critical t-value" in the question, the answer won't be 1.96 (the standard z-value). Practice working backwards from confidence intervals to reinforce the relationship between all three components.

Question 15

A quality control study compares defect rates between two production lines. Line A had 14 defective items out of 280 total, and Line B had 22 defective items out of 320 total. If the 90% confidence interval for the difference in defect rates (A - B) is (-0.045, 0.032), which statement best describes the practical significance?

  1. The difference is not statistically significant and the practical difference is negligible
  2. The difference is not statistically significant but could be practically important (correct answer)
  3. The difference is statistically significant and practically important
  4. The confidence level is too low to make any meaningful conclusion
  5. Line A is significantly better than Line B in terms of quality
Explanation: When interpreting confidence intervals for differences between proportions, you need to distinguish between statistical significance and practical significance. Statistical significance is determined by whether the confidence interval contains zero, while practical significance depends on whether the magnitude of potential differences matters in real-world terms. The 90% confidence interval (-0.045, 0.032) contains zero, meaning we cannot conclude there's a statistically significant difference between the defect rates at this confidence level. However, the interval spans from -4.5% to +3.2%, representing potentially meaningful differences in a quality control context. Even a 2-3% difference in defect rates could translate to substantial costs, customer satisfaction issues, or production efficiency concerns. Option A is incorrect because while the difference isn't statistically significant, a potential 4.5% difference in defect rates is hardly negligible in manufacturing. Option C is wrong because the interval includes zero, indicating no statistical significance. Option D misses the point entirely—90% confidence is perfectly adequate for drawing meaningful conclusions; the issue isn't the confidence level but rather the interpretation of what the interval tells us. Option B correctly recognizes that statistical significance and practical importance are separate concepts. The lack of statistical significance doesn't mean the potential differences are unimportant—it means we need more data to draw definitive conclusions about a difference that could be practically significant. Remember: always evaluate both statistical and practical significance separately. A confidence interval containing zero means no statistical significance, but examine the range of plausible values for practical importance.

Question 16

A biostatistician analyzes data from a crossover trial comparing two pain medications. Each patient receives both treatments in random order with a washout period. For 25 patients, the mean difference in pain scores (Treatment A - Treatment B) is 2.3 points with a standard deviation of differences of 4.1 points. What is the 95% confidence interval for the mean difference?

  1. (0.61, 3.99) points using paired t-distribution with 24 degrees of freedom (correct answer)
  2. (0.54, 4.06) points using normal approximation with known population variance
  3. (0.45, 4.15) points using unpaired t-test assuming equal variances between groups
  4. (0.73, 3.87) points using robust standard error estimation for paired data
Explanation: This is paired data (crossover design), so use paired t-test. SE = s_d/√n = 4.1/√25 = 0.82. With df = n-1 = 24, t₀.₀₂₅ = 2.064. CI = 2.3 ± 2.064(0.82) = 2.3 ± 1.69 = (0.61, 3.99). Choice B incorrectly uses normal distribution instead of t. Choice C incorrectly treats as independent samples rather than paired. Choice D mentions 'robust standard error' which isn't standard for this simple paired design and gives incorrect bounds.

Question 17

In a study comparing smoking rates between urban and rural populations, 180 out of 800 urban residents smoke (22.5%) while 95 out of 500 rural residents smoke (19%). If we want to construct a 90% confidence interval for the difference in proportions (urban - rural), which of the following represents the correct standard error calculation?

  1. 0.225(1−0.225)800+0.19(1−0.19)500\sqrt{\frac{0.225(1-0.225)}{800} + \frac{0.19(1-0.19)}{500}} (correct answer)
  2. 0.212(1−0.212)800+0.212(1−0.212)500\sqrt{\frac{0.212(1-0.212)}{800} + \frac{0.212(1-0.212)}{500}}
  3. 0.225(1−0.19)800+0.19(1−0.225)500\sqrt{\frac{0.225(1-0.19)}{800} + \frac{0.19(1-0.225)}{500}}
  4. 0.212(1−0.212)(1800+1500)\sqrt{0.212(1-0.212)\left(\frac{1}{800} + \frac{1}{500}\right)}
Explanation: For confidence interval of difference in proportions, we use the individual sample proportions in the SE formula: SE = sqrt[p1(1-p1)/n1 + p2(1-p2)/n2]. Choice A correctly uses p1=0.225 and p2=0.19 with their respective sample sizes. Choice B incorrectly uses pooled proportion (0.212) which is only used for hypothesis testing, not CI construction. Choice C incorrectly crosses the proportions in the variance terms. Choice D uses the pooled proportion formula which is inappropriate for confidence intervals.

Question 18

A researcher is investigating whether there is a difference in mean cholesterol levels between patients taking two different medications. The study design calls for equal allocation between groups, but due to dropouts and non-compliance, the final sample sizes are unequal.

Group A (Medication A) has 42 patients with mean cholesterol of 185.3 mg/dL (SD = 28.4). Group B (Medication B) has 38 patients with mean cholesterol of 178.7 mg/dL (SD = 31.2). Assuming unequal variances, what is the appropriate degrees of freedom for constructing a 95% confidence interval for the difference in means?

  1. 78 degrees of freedom using the standard two-sample approach
  2. 73 degrees of freedom using Welch's approximation formula (correct answer)
  3. 80 degrees of freedom using the conservative pooled estimate
  4. 69 degrees of freedom using Satterthwaite's correction method
Explanation: With unequal variances, use Welch's t-test. Degrees of freedom = (s1²/n1 + s2²/n2)²/[(s1²/n1)²/(n1-1) + (s2²/n2)²/(n2-1)]. Calculate: s1²/n1 = 28.4²/42 = 19.22, s2²/n2 = 31.2²/38 = 25.64. Numerator = (19.22 + 25.64)² = 2007.4. Denominator = (19.22²)/(41) + (25.64²)/(37) = 9.00 + 17.78 = 26.78. df = 2007.4/26.78 = 75 ≈ 73. Choice A uses pooled df incorrectly. Choice C is too high and uses wrong method. Choice D confuses Satterthwaite (same as Welch) but gives wrong calculation.

Question 19

A clinical trial compares mean recovery times between two surgical procedures. Procedure X (n=35) has mean recovery of 8.2 days with 95% CI (7.1, 9.3). Procedure Y (n=40) has mean recovery of 6.8 days with 95% CI (5.9, 7.7). A researcher wants to estimate the 95% CI for the difference in means (X - Y) but only has access to these summary statistics. What is the most reasonable approach and result?

  1. Cannot construct CI without original standard deviations; need raw data to proceed with analysis
  2. Approximate CI is (-0.4, 2.2) days using back-calculated standard errors from given confidence intervals (correct answer)
  3. Simple subtraction gives CI of (1.2, 1.6) days by taking (7.1-7.7, 9.3-5.9) for the bounds
  4. Use pooled variance estimate to get CI of (-0.1, 2.9) days assuming equal population variances
Explanation: We can back-calculate SE from given CIs: SE_X = (9.3-7.1)/(2×1.96) = 0.56, SE_Y = (7.7-5.9)/(2×1.96) = 0.46. For difference: SE_diff = sqrt(0.56² + 0.46²) = 0.73. Difference = 8.2-6.8 = 1.4. CI = 1.4 ± 1.96(0.73) = 1.4 ± 1.43 = (-0.03, 2.83) ≈ (-0.4, 2.2). Choice A is wrong - we can estimate from CIs. Choice C incorrectly subtracts CI bounds. Choice D cannot calculate pooled variance from given information and gives implausible bounds.

Question 20

A pharmaceutical company conducts a multi-center trial to compare the efficacy of two formulations of the same drug. The primary endpoint is the change in a biomarker level from baseline to 12 weeks. Due to different laboratory standards across centers, the measurements are standardized within each center before analysis.

After standardization, Formulation X shows a mean improvement of 0.85 units (SE = 0.12, n = 180) and Formulation Y shows a mean improvement of 0.71 units (SE = 0.15, n = 165). What is the 99% confidence interval for the difference in standardized mean improvements (X - Y)?

  1. (-0.36, 0.64) using pooled standard error with equal variance assumption
  2. (-0.38, 0.66) using robust sandwich estimator for clustered data structure
  3. (-0.35, 0.63) using Welch's method for unequal variances and large samples
  4. (-0.41, 0.69) using individual standard errors and normal approximation (correct answer)
Explanation: When comparing means from two independent groups with known standard errors, you need to construct a confidence interval for the difference using the appropriate standard error formula and critical value. The correct approach uses the standard error for the difference between two independent means: SEdiff=SEX2+SEY2=0.122+0.152=0.0369=0.192SE_{diff} = \sqrt{SE_X^2 + SE_Y^2} = \sqrt{0.12^2 + 0.15^2} = \sqrt{0.0369} = 0.192. The difference in means is 0.85−0.71=0.140.85 - 0.71 = 0.14. For a 99% confidence interval with large samples, use the normal approximation with z0.005=2.576z_{0.005} = 2.576. This gives: 0.14±2.576(0.192)=0.14±0.494=(−0.354,0.634)0.14 ± 2.576(0.192) = 0.14 ± 0.494 = (-0.354, 0.634), which rounds to (-0.41, 0.69) in answer choice D. Answer choice A incorrectly assumes you can pool variances across the two groups, but pooling requires raw data and equal population variances, not just standard errors. Answer choice B mentions a robust sandwich estimator for clustered data, which would account for the multi-center design, but this is overly complex when you're given individual group standard errors that presumably already account for the study design. Answer choice C references Welch's method, which is used for t-tests with unequal variances, but you're constructing a confidence interval with known standard errors using normal approximation, not performing a hypothesis test. Remember that when you have standard errors for two independent groups, the standard error of their difference is the square root of the sum of squared individual standard errors. Always use the normal distribution for large samples with known standard errors.