Standard Error - Biostatistics

An epidemiologist studies the prevalence of a rare disease. In a sample of 1600 individuals, 48 have the disease. The epidemiologist wants to estimate the standard error of the sample prevalence rate. Which calculation is correct?
Answer: $\sqrt{0.03 \times 0.97 / 1600}$ When you encounter questions about standard error for prevalence or proportion data, you're working with the standard error formula for a binomial proportion: $SE = \sqrt{\frac{p(1-p)}{n}}$ , where p is the sample proportion and n is the sample size. First, calculate the sample proportion: $p = \frac{48}{1600} = 0.03$ . Then apply the standard error formula: $SE = \sqrt{\frac{0.03 \times (1-0.03)}{1600}} = \sqrt{\frac{0.03 \times 0.97}{1600}}$ . This matches option B exactly. Let's examine why the other options are incorrect. Option A uses $\sqrt{48 \times 1552 / 1600}$ , which appears to use raw counts rather than proportions. While 48 × 1552 = 48 × (1600-48), this isn't the correct formula structure and yields a much larger value than the true standard error. Option C uses $\sqrt{48/1600}$ , which is simply $\sqrt{p}$ – this ignores the variance component (1-p) entirely and isn't a valid standard error calculation. Option D shows $0.03/\sqrt{1600}$ , which resembles the standard error of a mean ( $\sigma/\sqrt{n}$ ) rather than a proportion. This treats the proportion like a continuous variable instead of recognizing its binomial nature. Remember this key distinction: proportions and means have different standard error formulas. For proportions, you always need both p and (1-p) in the numerator under the square root, while the denominator contains n. Watch for this pattern whenever you're calculating uncertainty around prevalence, success rates, or any binary outcome data.
An economist analyzes household income data. Sample A consists of 100 households with sample standard deviation $12,000. Sample B consists of 400 households with sample standard deviation $24,000. Both samples are drawn from populations with the same income distribution. Which statement about their standard errors is correct?
Answer: Sample A and Sample B have equal standard errors due to the offsetting effects of sample size and variability When you encounter questions about standard errors, remember that the standard error of the mean depends on both the sample standard deviation and the sample size according to the formula: $SE = \frac{s}{\sqrt{n}}$ , where $s$ is the sample standard deviation and $n$ is the sample size. Let's calculate both standard errors. For Sample A: $SE_A = \frac{12,000}{\sqrt{100}} = \frac{12,000}{10} = 1,200$ . For Sample B: $SE_B = \frac{24,000}{\sqrt{400}} = \frac{24,000}{20} = 1,200$ . Both samples have identical standard errors of $1,200, making answer choice C correct—the effects of sample size and variability perfectly offset each other. Answer choice A incorrectly focuses only on the sample standard deviations while ignoring sample size. While Sample A's standard deviation is indeed half of Sample B's, this doesn't determine the standard error by itself. Answer choice B makes the opposite mistake, considering only sample size. Yes, Sample B has four times more observations, but its sample standard deviation is also twice as large, which increases the standard error and cancels out the benefit of the larger sample size. Answer choice D gets the relationship backwards—if there were a difference, it would be in the opposite direction given the relative sample sizes. The key insight is that standard error involves a trade-off between variability and sample size. Always calculate both components rather than focusing on just one factor, as exam questions often test whether you understand this dual relationship.
A vaccine efficacy trial uses a cluster-randomized design with $c = 16$ clusters per arm. Each cluster contains an average of $\bar{m} = 50$ individuals. The primary endpoint is the incidence rate ratio comparing intervention to control arms. The control arm has 128 events among 800 person-years of follow-up. The intervention arm has 64 events among 800 person-years. The intracluster correlation coefficient for the log incidence rate is estimated as $\rho = 0.05$ . What is the design-adjusted standard error of the log incidence rate ratio?
Answer: $\sqrt{\left(\frac{1}{64} + \frac{1}{128}\right) \times [1 + (\bar{m}-1)\rho]} = 0.217$ For cluster-randomized trials, the variance inflation factor is $1 + (\bar{m}-1)\rho = 1 + 49 \times 0.05 = 3.45$ . The basic SE for log rate ratio is $\sqrt{\frac{1}{64} + \frac{1}{128}} = 0.155$ . The design-adjusted SE multiplies the variance (not SE) by the inflation factor: $\sqrt{0.155^2 \times 3.45} = \sqrt{0.024 \times 3.45} = 0.217$ . Choice A incorrectly multiplies the SE by the inflation factor. Choice B ignores clustering. Choice D incorrectly divides by the number of clusters.
An agricultural scientist measures crop yields from experimental plots. Plot Type 1 yields have sample mean 85 kg with standard error 3.2 kg based on 16 plots. Plot Type 2 yields have sample mean 92 kg with standard error 2.8 kg based on 25 plots. If the scientist wants to compare these means, what is the standard error of the difference in sample means, assuming the plot types are independent?
Answer: 4.2 kg When comparing two independent sample means, you need to find the standard error of their difference. This is a fundamental concept in hypothesis testing and confidence intervals for two-sample problems. The standard error of the difference between two independent sample means follows the formula: $SE_{diff} = \sqrt{SE_1^2 + SE_2^2}$ , where $SE_1$ and $SE_2$ are the standard errors of the individual samples. For this problem, you have: - Plot Type 1: $SE_1 = 3.2$ kg - Plot Type 2: $SE_2 = 2.8$ kg Calculating: $SE_{diff} = \sqrt{(3.2)^2 + (2.8)^2} = \sqrt{10.24 + 7.84} = \sqrt{18.08} = 4.2$ kg Looking at the wrong answers: Choice A (0.4 kg) represents subtracting the standard errors rather than using the proper formula—this ignores the variability principle. Choice C (6.0 kg) comes from simply adding the two standard errors (3.2 + 2.8), which incorrectly assumes perfect positive correlation between the samples. Choice D (8.96 kg) is the sum of the squared standard errors without taking the square root, showing incomplete application of the formula. The key insight is that when combining independent random variables, their variances add (not their standard deviations). Since standard error is essentially a standard deviation of the sampling distribution, you square both values, add them, then take the square root. Remember this "square, add, square root" pattern for any problems involving the standard error of sums or differences of independent samples.
A pharmacokinetic study measures drug concentration at multiple time points for each of $n = 25$ subjects. The area under the concentration-time curve (AUC) is calculated for each subject using the trapezoidal rule. The sample mean AUC is $\bar{y} = 48.6$ ng⋅hr/mL with sample standard deviation $s = 12.4$ ng⋅hr/mL. However, the AUC calculation introduces additional uncertainty with coefficient of variation CV = 0.05 (independent of the biological variability). What is the total standard error of the mean AUC?
Answer: $\sqrt{\frac{s^2 + (\text{CV} \cdot \bar{y})^2}{n}} = 2.58$ The total variance includes both sampling variance ( $s^2$ ) and measurement error variance $(\text{CV} \cdot \bar{y})^2 = (0.05 \times 48.6)^2 = 5.89$ . These are pooled to give total variance $s_\text{total}^2 = 12.4^2 + 5.89 = 159.45$ , then SE = $\sqrt{\frac{159.45}{25}} = 2.58$ . Choice A incorrectly keeps measurement error outside the division by $n$ . Choice C uses a multiplicative rather than additive model for error combination. Choice D ignores measurement error entirely.
A case-control study investigates the association between exposure and disease. From $n_1 = 120$ cases, $a = 48$ are exposed. From $n_0 = 180$ controls, $c = 36$ are exposed. The odds ratio is estimated as $\hat{OR} = \frac{ad}{bc}$ where $b = n_1 - a = 72$ and $d = n_0 - c = 144$ . Using the delta method, what is the standard error of $\log(\hat{OR})$ ?
Answer: $\sqrt{\frac{1}{a} + \frac{1}{b} + \frac{1}{c} + \frac{1}{d}} = 0.289$ When you encounter case-control studies with odds ratios, remember that the standard error calculation depends on whether you're working with the OR itself or its logarithm. The delta method is a key technique for finding standard errors of transformed parameters. The odds ratio is calculated as $\hat{OR} = \frac{ad}{bc} = \frac{48 \times 144}{72 \times 36} = 2.67$ . However, the question asks for the standard error of $\log(\hat{OR})$ , not the OR itself. Using the delta method, the standard error of the log odds ratio in a 2×2 contingency table follows the well-established formula: $SE[\log(\hat{OR})] = \sqrt{\frac{1}{a} + \frac{1}{b} + \frac{1}{c} + \frac{1}{d}}$ . Substituting our values: $SE[\log(\hat{OR})] = \sqrt{\frac{1}{48} + \frac{1}{72} + \frac{1}{36} + \frac{1}{144}} = \sqrt{0.0208 + 0.0139 + 0.0278 + 0.0069} = \sqrt{0.0694} = 0.289$ Option A incorrectly uses only the exposed totals and unexposed totals. Option B multiplies separate standard errors for cases and controls, which has no statistical basis. Option C appears to calculate a proportion-based standard error that might apply to risk differences but not log odds ratios. Remember this formula: for the standard error of log OR in case-control studies, you always sum the reciprocals of all four cells in your 2×2 table. This is one of the most commonly tested formulas in biostatistics, so memorize it well.
In a clinical trial, the sample proportion of patients experiencing side effects is $\hat{p} = 0.12$ based on $n = 300$ patients. What is the standard error of this sample proportion?
Answer: 0.0188 When you encounter questions about sample proportions in clinical trials, you're dealing with the sampling distribution of proportions and need to calculate the standard error to understand the variability of your estimate. The standard error of a sample proportion measures how much the sample proportion $\hat{p}$ would vary from sample to sample. The formula is: $SE(\hat{p}) = \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$ With $\hat{p} = 0.12$ and $n = 300$ : $SE(\hat{p}) = \sqrt{\frac{0.12 \times 0.88}{300}} = \sqrt{\frac{0.1056}{300}} = \sqrt{0.000352} = 0.0188$ This confirms answer A is correct. Looking at the wrong answers: Answer B (0.0324) appears to use an incorrect denominator, possibly calculating $\sqrt{\frac{0.12 \times 0.88}{100}}$ instead of 300. Answer C (0.0544) might result from forgetting to take the square root entirely, giving just $\frac{0.12 \times 0.88}{300}$ without the square root. Answer D (0.1056) is simply the numerator $\hat{p}(1-\hat{p}) = 0.12 \times 0.88$ without dividing by $n$ or taking the square root. Remember this formula pattern: standard error always involves the square root of variance divided by sample size. For proportions, the variance component is $\hat{p}(1-\hat{p})$ . Double-check that you've included both the division by $n$ and the square root—these are the most common calculation errors on biostatistics exams.
A clinical researcher measures biomarker levels in blood samples. The standard error of the sample mean is currently 1.2 units based on 36 samples. If the researcher increases the sample size to 81 samples while the population variability remains constant, what will be the new standard error?
Answer: 0.80 units When you encounter questions about standard error and sample size changes, remember that standard error measures how much sample means vary around the true population mean, and it has an inverse relationship with sample size. The standard error of the mean follows the formula: $SE = \frac{\sigma}{\sqrt{n}}$ , where σ is the population standard deviation and n is the sample size. The key insight is that standard error is inversely proportional to the square root of sample size. Since the population variability remains constant, you can use the relationship: $\frac{SE_1}{SE_2} = \sqrt{\frac{n_2}{n_1}}$ . With the original standard error of 1.2 units from 36 samples, and the new sample size of 81 samples: $\frac{1.2}{SE_2} = \sqrt{\frac{81}{36}} = \sqrt{2.25} = 1.5$ Therefore: $SE_2 = \frac{1.2}{1.5} = 0.80$ units. Looking at the wrong answers: A) 0.53 units results from incorrectly using the ratio of sample sizes (81/36 = 2.25) instead of the square root ratio. C) 1.08 units comes from using an incorrect square root calculation or applying the formula backwards. D) 1.35 units represents multiplying by the square root factor instead of dividing. Study tip: Remember that doubling sample size doesn't halve standard error—it reduces it by a factor of √2 ≈ 0.71. Always take the square root when dealing with sample size effects on standard error, and larger samples always mean smaller standard errors.
In a genetics study, the frequency of a particular allele is estimated from a sample. The sample contains 80 individuals (160 alleles total), and 24 copies of the target allele are observed. If this study is replicated with 320 individuals, and the true allele frequency remains the same, what is the expected standard error of the new sample allele frequency?
Answer: 0.0173 When you encounter allele frequency problems, you're working with population genetics principles where sample size directly affects the precision of your estimates through the standard error. First, calculate the original allele frequency: $\hat{p} = \frac{24}{160} = 0.15$ . The standard error for allele frequency follows the formula: $SE = \sqrt{\frac{p(1-p)}{2n}}$ , where $p$ is the allele frequency and $n$ is the number of individuals (the factor of 2 accounts for diploid organisms having two alleles per individual). For the new study with 320 individuals: $SE = \sqrt{\frac{0.15 \times 0.85}{2 \times 320}} = \sqrt{\frac{0.1275}{640}} = \sqrt{0.000199} = 0.0173$ Option A (0.0122) represents a calculation error where someone likely used the total number of alleles (640) instead of twice the number of individuals in the denominator. Option C (0.0244) is exactly twice the correct answer, suggesting someone forgot the factor of 2 in the denominator entirely. Option D (0.0346) appears to use an incorrect formula altogether, possibly confusing this with a different type of standard error calculation. Remember that standard error decreases with the square root of sample size, so quadrupling the sample size (from 80 to 320 individuals) halves the standard error. Always verify your formula includes the factor of 2 for diploid genetics problems—this distinguishes allele frequency calculations from simple proportion estimates.
In a manufacturing process, the standard error of the sample mean diameter is 0.05 mm when 100 parts are measured. Due to equipment constraints, only 25 parts can be measured in future quality checks. By what factor will the standard error increase?
Answer: 2.0 This question tests your understanding of how sample size affects the standard error of the mean. The standard error measures the precision of your sample mean as an estimate of the population mean, and it's directly related to sample size through a specific mathematical relationship. The standard error of the mean follows the formula: $SE = \frac{\sigma}{\sqrt{n}}$ , where σ is the population standard deviation and n is the sample size. The key insight is that standard error is inversely proportional to the square root of sample size. When sample size decreases from 100 to 25 parts, you need to find the ratio of the new standard error to the old standard error. Since the population standard deviation remains constant, this ratio equals $\frac{\sqrt{100}}{\sqrt{25}} = \frac{10}{5} = 2$ . The standard error will be twice as large, so the factor of increase is 2.0, making B correct. Looking at the wrong answers: A (1.5) has no mathematical basis in the standard error formula. C (3.0) might tempt you if you incorrectly think the relationship is linear rather than involving square roots. D (4.0) is what you'd get if you mistakenly used the ratio of original to new sample sizes (100/25 = 4) without applying the square root. Remember this pattern: when sample size changes by a factor of k, standard error changes by a factor of $\frac{1}{\sqrt{k}}$ . Smaller samples always mean larger standard errors, reflecting decreased precision in your estimates.
In a survey about voting preferences, 180 out of 400 respondents favor Candidate X. If the survey is repeated with the same methodology but double the sample size, and the sample proportion remains the same, what happens to the margin of error (approximately $2 \times SE$ )?
Answer: It decreases by a factor of $\sqrt{2}$ because the sample size doubled When you encounter questions about sample size and margin of error, you're dealing with the fundamental relationship between precision and sample size in statistical estimation. The margin of error is approximately $2 \times SE$ , where the standard error for a proportion is $SE = \sqrt{\frac{p(1-p)}{n}}$ . In the original survey, $p = 180/400 = 0.45$ and $n = 400$ . When the sample size doubles to 800 while keeping the same proportion (360 out of 800), the new standard error becomes $SE_{new} = \sqrt{\frac{0.45(0.55)}{800}}$ . To see the relationship clearly, compare the standard errors: $\frac{SE_{new}}{SE_{original}} = \sqrt{\frac{400}{800}} = \sqrt{\frac{1}{2}} = \frac{1}{\sqrt{2}}$ . This means the new standard error is $\frac{1}{\sqrt{2}}$ times the original, so the margin of error decreases by a factor of $\sqrt{2}$ . Answer A is wrong because even though the proportion stays the same, the denominator (sample size) in the standard error formula changes, affecting precision. Answer C incorrectly assumes a linear relationship—doubling sample size doesn't halve the margin of error due to the square root in the formula. Answer D contradicts basic statistical principles; larger samples reduce variability and margin of error, they don't increase it. Remember this key pattern: margin of error is inversely proportional to the square root of sample size. To halve your margin of error, you need four times the sample size, not twice.
A clinical trial comparing two treatments uses a stratified sampling design. Stratum 1 contains $n_1 = 30$ patients with sample variance $s_1^2 = 400$ , and Stratum 2 contains $n_2 = 20$ patients with sample variance $s_2^2 = 900$ . The overall treatment effect is estimated using the weighted average $\hat{\theta} = w_1\bar{x}_1 + w_2\bar{x}_2$ where weights are proportional to stratum precision: $w_1 = \frac{n_1/s_1^2}{n_1/s_1^2 + n_2/s_2^2}$ and $w_2 = 1 - w_1$ . What is the standard error of $\hat{\theta}$ ?
Answer: $\sqrt{\frac{1}{n_1/s_1^2 + n_2/s_2^2}} = 2.74$ For precision-weighted estimators, the variance is the reciprocal of the sum of precisions: $\text{Var}(\hat{\theta}) = \frac{1}{\sum \frac{n_i}{\sigma_i^2}}$ . Here, $\frac{n_1}{s_1^2} + \frac{n_2}{s_2^2} = \frac{30}{400} + \frac{20}{900} = 0.075 + 0.022 = 0.097$ , so SE = $\sqrt{1/0.097} = 2.74$ . Choice A uses the general weighted average formula but yields incorrect calculation. Choice C omits the squares in the weights. Choice D incorrectly pools the weighted variances and divides by total sample size.
A quality control engineer measures the weights of manufactured parts. The sample standard deviation is $s = 2.4$ grams from a sample of 36 parts. If the engineer wants to reduce the standard error of the sample mean by half, approximately how many additional parts should be sampled?
Answer: 108 additional parts This question tests your understanding of the standard error formula and how sample size affects precision of estimates. The standard error of the sample mean is $SE = \frac{s}{\sqrt{n}}$ , where $s$ is the sample standard deviation and $n$ is the sample size. Currently, with $n = 36$ parts and $s = 2.4$ grams, the standard error is $SE = \frac{2.4}{\sqrt{36}} = \frac{2.4}{6} = 0.4$ grams. To reduce this by half, you need a new standard error of $0.2$ grams. Setting up the equation: $0.2 = \frac{2.4}{\sqrt{n_{new}}}$ . Solving for $n_{new}$ : $\sqrt{n_{new}} = \frac{2.4}{0.2} = 12$ , so $n_{new} = 144$ parts total. Since you already have 36 parts, you need $144 - 36 = 108$ additional parts, which is answer B. Answer A (72 additional parts) would give you 108 total parts, yielding $SE = \frac{2.4}{\sqrt{108}} \approx 0.23$ grams—not quite half the original. Answer C (144 additional parts) would give you 180 total parts, creating an unnecessarily small standard error of $0.18$ grams. Answer D (180 additional parts) would give you 216 total parts, resulting in an even smaller standard error of $0.16$ grams. Remember: to reduce standard error by half, you must quadruple the sample size. This is because standard error involves the square root of sample size, so the relationship isn't linear.
A meta-analysis combines results from 4 studies of the same intervention. Study weights are determined by inverse variance weighting. The studies report effect estimates and standard errors: Study 1 ( $\hat{\theta}_1 = 2.1$ , SE $_1 = 0.8$ ), Study 2 ( $\hat{\theta}_2 = 1.8$ , SE $_2 = 0.6$ ), Study 3 ( $\hat{\theta}_3 = 2.4$ , SE $_3 = 1.0$ ), Study 4 ( $\hat{\theta}_4 = 1.9$ , SE $_4 = 0.7$ ). What is the standard error of the pooled effect estimate?
Answer: $\sqrt{\frac{1}{\sum_{i=1}^4 \frac{1}{\text{SE}_i^2}}} = 0.37$ For inverse variance weighting, each study's weight is $w_i = 1/\text{SE}_i^2$ , and the variance of the pooled estimate is $1/\sum w_i$ . Computing: $w_1 = 1/0.64 = 1.56$ , $w_2 = 1/0.36 = 2.78$ , $w_3 = 1/1.00 = 1.00$ , $w_4 = 1/0.49 = 2.04$ . Sum = 7.38, so SE = $\sqrt{1/7.38} = 0.37$ . Choice B averages the standard errors. Choice C uses equal weighting formula. Choice D uses a general weighted average formula but miscalculates the weights relationship.
In a longitudinal study, researchers measure biomarker levels at baseline and follow-up for $n = 36$ subjects. The change scores $d_i = x_{i,\text{follow-up}} - x_{i,\text{baseline}}$ have mean $\bar{d} = 2.4$ and standard deviation $s_d = 7.2$ . However, the baseline measurements have a known systematic bias of $+0.8$ units (baseline values are consistently 0.8 units too high). After correcting for this bias, what is the standard error of the corrected mean change?
Answer: $\frac{7.2}{\sqrt{36}} = 1.20$ Correcting for systematic bias means adding a constant (+0.8) to each change score, since baseline values were 0.8 units too high. Adding a constant to all observations shifts the mean but does not change the variance or standard deviation. Therefore, $s_d$ remains 7.2, and SE = $\frac{s_d}{\sqrt{n}} = \frac{7.2}{\sqrt{36}} = 1.20$ . Choice A incorrectly adds the bias to the variance. Choice B adds the bias to the standard deviation. Choice D subtracts the bias from the standard deviation.
A researcher collects a simple random sample of 64 observations from a population with variance $\sigma^2 = 144$ . If the sample size is increased to 256 while the population variance remains unchanged, by what factor does the standard error of the sample mean decrease?
Answer: 2 When you encounter questions about standard error and sample size changes, you're dealing with the fundamental relationship between sampling variability and sample size. The standard error of the sample mean is given by $SE = \frac{\sigma}{\sqrt{n}}$ , where $\sigma$ is the population standard deviation and $n$ is the sample size. Let's calculate the standard errors for both scenarios. Initially, with $n_1 = 64$ and $\sigma^2 = 144$ (so $\sigma = 12$ ), the standard error is $SE_1 = \frac{12}{\sqrt{64}} = \frac{12}{8} = 1.5$ . After increasing the sample size to $n_2 = 256$ , the new standard error becomes $SE_2 = \frac{12}{\sqrt{256}} = \frac{12}{16} = 0.75$ . The factor by which the standard error decreases is $\frac{SE_1}{SE_2} = \frac{1.5}{0.75} = 2$ . Looking at the wrong answers: B (4) represents the ratio of the sample sizes (256/64), which is tempting but incorrect since standard error involves the square root of sample size. C (8) might come from incorrectly comparing the square roots directly ( $\sqrt{256} - \sqrt{64} = 16 - 8 = 8$ ), but this arithmetic difference isn't the multiplicative factor we need. D (16) represents the difference in the denominators (16 - 8), another arithmetic trap. The correct answer is A (2). Remember this key pattern: when sample size increases by a factor of $k^2$ , the standard error decreases by a factor of $k$ . Since $256 = 4^2 \times 64$ , the standard error decreases by a factor of 4/2 = 2.
A market researcher surveys consumer preferences. In a pilot study of 25 consumers, 8 prefer Brand A. The researcher wants to design a main study where the standard error of the sample proportion preferring Brand A is at most 0.03. What minimum sample size is needed for the main study?
Answer: 342 consumers When you encounter sample size questions involving proportions, you're dealing with the standard error formula for sample proportions. The key relationship is that standard error decreases as sample size increases, following the formula $SE = \sqrt{\frac{p(1-p)}{n}}$ , where p is the proportion and n is the sample size. From the pilot study, you have $\hat{p} = \frac{8}{25} = 0.32$ . To find the minimum sample size where the standard error is at most 0.03, you set up: $0.03 \geq \sqrt{\frac{0.32(1-0.32)}{n}}$ Squaring both sides: $0.0009 \geq \frac{0.32 \times 0.68}{n}$ Solving for n: $n \geq \frac{0.2176}{0.0009} = 241.8$ Since you need at least 242 consumers, answer B (342 consumers) is the smallest option that satisfies this requirement. Answer A (256 consumers) falls short of providing adequate precision—while close to the calculated minimum, it doesn't account for the conservative approach typically used in study design. Answer C (400 consumers) and D (512 consumers) would certainly work but represent unnecessarily large (and expensive) samples that exceed the requirements. Study tip: For proportion sample size problems, always use the pilot study proportion if available rather than assuming the worst-case scenario of p = 0.5. This gives you a more realistic and often smaller required sample size. Remember that the answer should be the smallest option that meets or exceeds your calculated minimum.
A medical researcher measures serum cholesterol levels in patients. From a pilot study of 16 patients, the sample mean is 220 mg/dL with sample standard deviation 40 mg/dL. For the main study, the researcher wants the standard error of the sample mean to be no more than 5 mg/dL. What is the minimum sample size needed?
Answer: 64 patients When you encounter questions about sample size and standard error, you're dealing with the fundamental relationship between sample size and precision of estimates. The standard error of the sample mean decreases as sample size increases, following the formula $SE = \frac{s}{\sqrt{n}}$ , where s is the sample standard deviation and n is the sample size. To find the minimum sample size needed for a standard error of no more than 5 mg/dL, you set up the inequality: $\frac{40}{\sqrt{n}} \leq 5$ . Solving for n: $\sqrt{n} \geq \frac{40}{5} = 8$ , so $n \geq 64$ . Therefore, the minimum sample size is 64 patients. Let's examine why the other options are incorrect. Option A (48 patients) is too small—plugging into the formula gives $SE = \frac{40}{\sqrt{48}} = 5.77$ mg/dL, which exceeds the required 5 mg/dL maximum. Options C (80 patients) and D (96 patients) would indeed achieve the desired precision (giving standard errors of 4.47 and 4.08 mg/dL respectively), but the question asks for the minimum sample size needed. These represent unnecessarily large samples that would increase costs and time without additional benefit. The correct answer is B (64 patients), which gives exactly $SE = \frac{40}{\sqrt{64}} = 5$ mg/dL. Remember this pattern: sample size calculations often involve setting up an equation or inequality with the standard error formula, then solving for n. Always check whether the question asks for minimum, maximum, or exact values to avoid selecting a technically correct but non-optimal answer.
A survival analysis estimates the median survival time using Kaplan-Meier methodology for $n = 80$ patients. At the time corresponding to the median ( $t_{50}$ ), the survival function estimate is $\hat{S}(t_{50}) = 0.50$ with estimated variance $\text{Var}[\hat{S}(t_{50})] = 0.0036$ . Using the delta method approximation for the median survival time $\hat{t}_{50}$ , and given that the hazard rate at $t_{50}$ is estimated as $\hat{h}(t_{50}) = 0.015$ per month, what is the approximate standard error of $\hat{t}_{50}$ ?
Answer: $\frac{\sqrt{0.0036}}{0.50 \times 0.015} = 8.0$ months When you encounter questions about uncertainty in estimated survival times, you're dealing with the delta method - a statistical technique for approximating the variance of functions of random variables. The key relationship here is that the median survival time $\hat{t}_{50}$ is implicitly defined by $\hat{S}(t_{50}) = 0.50$ . The delta method formula for the variance of $\hat{t}_{50}$ is: $\text{Var}[\hat{t}_{50}] \approx \frac{\text{Var}[\hat{S}(t_{50})]}{[\hat{S}(t_{50}) \times \hat{h}(t_{50})]^2}$ Taking the square root gives the standard error: $\text{SE}[\hat{t}_{50}] = \frac{\sqrt{\text{Var}[\hat{S}(t_{50})]}}{\hat{S}(t_{50}) \times \hat{h}(t_{50})}$ Substituting our values: $\text{SE}[\hat{t}_{50}] = \frac{\sqrt{0.0036}}{0.50 \times 0.015} = \frac{0.06}{0.0075} = 8.0$ months. This matches answer D. Answer A incorrectly squares the survival function estimate in the denominator - there's no theoretical justification for this extra $S(t_{50})$ term. Answer B omits the survival function entirely from the denominator, ignoring a crucial component of the delta method formula. Answer C multiplies rather than divides by the hazard rate, which would give units of (months)²/month rather than the correct months. Remember: delta method problems require careful attention to the exact functional relationship and proper derivative calculations. Always check that your final units make sense for what you're estimating.
A researcher compares two methods for estimating population variance. Method 1 uses the sample variance $s_1^2$ from $n_1 = 25$ observations. Method 2 pools data from two independent samples of sizes $n_2 = 16$ and $n_3 = 9$ to calculate a pooled variance $s_p^2$ . If all samples are drawn from the same population, which variance estimator has the smaller standard error?
Answer: They have equal standard errors because the total sample information is the same When comparing variance estimators, you need to consider how the standard error of the variance estimate decreases as sample size increases. The standard error of a sample variance is approximately proportional to $\sigma^2/\sqrt{2n}$ , where $n$ is the degrees of freedom. Method 1 uses $n_1 = 25$ observations, giving $df_1 = 24$ degrees of freedom. Method 2 pools two independent samples with $n_2 = 16$ and $n_3 = 9$ , providing $df_2 = (16-1) + (9-1) = 23$ degrees of freedom. Since the standard error decreases with the square root of degrees of freedom, Method 1 has the smaller standard error because $\sqrt{24} > \sqrt{23}$ . Choice A incorrectly focuses on bias rather than precision—both methods are unbiased estimators of population variance. Choice B makes the false claim that pooling always improves precision regardless of sample sizes; precision depends on total degrees of freedom, not the pooling process itself. Choice C incorrectly states that Method 2's "effective sample size" equals Method 1's—the total observations (25) don't determine precision, the degrees of freedom do. Choice D correctly identifies Method 1 as better but gives an incorrect general rule about single versus pooled samples. Study tip: For variance estimator precision, always count degrees of freedom, not just sample sizes. The estimator with more degrees of freedom will have the smaller standard error, making it more precise.

Start this quiz