Sample Size Formulas for Common Experimental Designs
By Dr. Zubair Khalid, DVM, MS, PhD ·
Quick Answer
- Sample size formulas depend on the experimental design, the outcome variable type, and the statistical test planned for analysis.
- Select the formula that matches your design first, then input the expected effect size, variability, significance level, and desired power.
- All formulas produce estimates that assume correct inputs, so inaccurate guesses about variability or effect size will lead to underpowered or wasteful studies.
Understanding Sample Size in Experimental Design
Sample size determination is a foundational step in experimental design across the life sciences. The number of experimental units in a study directly affects the ability to detect true effects, the precision of estimates, and the ethical justification for using animals or human participants. Researchers in biology, laboratory science, and related fields must understand how sample size formulas work to design studies that are both efficient and scientifically valid.
The core principle is that sample size calculations balance two competing concerns. A sample that is too small may fail to detect a real biological effect, producing a false negative result. A sample that is too large wastes resources, exposes more subjects than necessary to experimental procedures, and may detect trivial effects that have no biological meaning. The goal is to calculate the minimum number of observations needed to achieve a specified probability of detecting an effect of a given size.
Statistical power is the probability that a study will detect an effect when one truly exists. Power depends on four interconnected elements: the significance level (alpha), the sample size, the effect size, and the variability of the data. When any three of these are fixed, the fourth can be calculated. Sample size formulas solve for the number of observations needed when the other three elements are specified.
The significance level, typically set at 0.05, represents the probability of rejecting the null hypothesis when it is actually true. This is the Type I error rate. Power, typically set at 0.80 or 0.90, represents the probability of correctly rejecting the null hypothesis when a true effect exists. This is the complement of the Type II error rate, which is the probability of failing to detect a real effect.
Effect size is the magnitude of the biological or experimental effect that the researcher expects to observe. This is often the most difficult input to specify because it requires prior knowledge from pilot studies, published literature, or theoretical reasoning. Variability, expressed as the standard deviation of the outcome measure, also requires prior knowledge or estimation.
The relationship between these elements is fundamental to all sample size calculations. Researchers must understand that sample size formulas do not produce a single correct answer. They produce an estimate based on the assumptions entered into the calculation. Different assumptions about effect size or variability will produce different sample size recommendations.
At a Glance
The following table summarizes the main sample size formulas covered in this article, the design type each applies to, and the key inputs required for each calculation.
| Design | Statistical Test | Key Inputs | Formula Type |
|---|---|---|---|
| One-sample mean | One-sample t-test | Effect size, standard deviation, alpha, power | n = (z_alpha/2 + z_beta)^2 * sigma^2 / delta^2 |
| Two independent groups | Two-sample t-test | Effect size, standard deviation, alpha, power | n per group = 2 (z_alpha/2 + z_beta)^2 sigma^2 / delta^2 |
| Multiple groups | One-way ANOVA | Number of groups, effect size, alpha, power | n per group = 2 (z_alpha/2 + z_beta)^2 sigma^2 / delta^2 (approximation) |
| Proportion | One-sample proportion test | Expected proportion, null proportion, alpha, power | n = (z_alpha/2 sqrt(p0(1-p0)) + z_beta sqrt(p1(1-p1)))^2 / (p1-p0)^2 |
| Two proportions | Two-sample proportion test | Proportions in each group, alpha, power | n per group = (z_alpha/2 sqrt(2p_bar(1-p_bar)) + z_beta sqrt(p1(1-p1) + p2(1-p2)))^2 / (p1-p2)^2 |
Core Principles of Sample Size Calculation
The Role of Statistical Power
Statistical power is the probability that a study will detect an effect of a specified magnitude when that effect truly exists in the population. Power is a critical consideration in experimental design because it directly determines the likelihood of obtaining a statistically significant result when a real biological difference is present.
The conventional target for power in life science research is 0.80, meaning that the study has an 80 percent chance of detecting the specified effect. Some studies, particularly those with high stakes or expensive data collection, may target 0.90 or higher. The choice of power level involves a tradeoff between the cost of collecting more samples and the risk of missing a real effect.
Power is not a fixed property of a study. It depends on the true effect size, the variability of the outcome measure, the sample size, and the significance level. Researchers can increase power by increasing the sample size, reducing measurement error, using a more sensitive outcome measure, or relaxing the significance level. Each of these changes has consequences for the interpretation of the results.
The relationship between power and sample size is nonlinear. Increasing the sample size from 10 to 20 per group produces a larger gain in power than increasing from 100 to 110 per group. This diminishing return means that researchers must consider the marginal benefit of each additional observation against the cost of collecting it.
Significance Level and Type I Error
The significance level, denoted as alpha, is the probability of rejecting the null hypothesis when it is true. This is the Type I error rate. The conventional alpha level in most life science research is 0.05, meaning that the researcher accepts a 5 percent chance of concluding that an effect exists when it does not.
The choice of alpha affects the sample size calculation. A smaller alpha, such as 0.01, requires a larger sample size to achieve the same power. This is because the critical value for the test statistic moves further into the tail of the distribution, making it harder to reject the null hypothesis.
When multiple comparisons are planned, the alpha level may need to be adjusted to control the overall Type I error rate. This adjustment, such as the Bonferroni correction, reduces the alpha for each individual test, which in turn increases the required sample size. Researchers should account for planned multiple comparisons when calculating sample size.
Effect Size and Its Estimation
Effect size is the magnitude of the difference or association that the researcher expects to observe. In the context of sample size calculation, effect size is typically expressed as the difference between the means of two groups divided by the standard deviation of the outcome variable. This standardized measure, often called Cohen's d, allows comparisons across studies with different measurement scales.
Estimating effect size is the most challenging part of sample size calculation because it requires prior knowledge. Researchers can obtain effect size estimates from pilot studies, published literature, or theoretical reasoning. When no prior information is available, researchers may use conventions for small, medium, and large effects, but these conventions are arbitrary and should be used with caution.
The effect size input has a direct impact on the calculated sample size. A larger effect size requires a smaller sample size to detect, while a smaller effect size requires a larger sample size. This relationship is inverse and nonlinear. Halving the effect size requires approximately quadrupling the sample size to maintain the same power.
Variability and Standard Deviation
The standard deviation of the outcome variable is a measure of the variability in the population. It is a critical input in sample size formulas because it determines the noise against which the effect must be detected. A larger standard deviation requires a larger sample size to achieve the same power.
The standard deviation can be estimated from pilot data, published literature, or prior studies in the same laboratory. When the standard deviation is unknown, researchers may use a conservative estimate that is larger than expected to ensure adequate power. This approach is safer than underestimating variability, which would produce a sample size that is too small.
The standard deviation is also affected by the measurement process. Reducing measurement error through standardized protocols, calibrated instruments, and trained personnel can reduce the standard deviation and therefore the required sample size. This is one reason why careful experimental technique is important for efficient study design.
One-Sample Designs
Formula for One-Sample Mean
The one-sample design is used when the researcher wants to compare the mean of a single group to a known or hypothesized value. This design is common in biology when testing whether a treatment produces a mean that differs from a control value or a theoretical expectation.
The sample size formula for a one-sample t-test is:
n = (z_alpha/2 + z_beta)^2 * sigma^2 / delta^2
In this formula, n is the required sample size, z_alpha/2 is the critical value from the standard normal distribution corresponding to the desired significance level, z_beta is the critical value corresponding to the desired power, sigma is the standard deviation of the outcome variable, and delta is the expected difference between the sample mean and the hypothesized value.
The z values are obtained from standard normal distribution tables or statistical software. For a significance level of 0.05, z_alpha/2 is 1.96. For power of 0.80, z_beta is 0.84. For power of 0.90, z_beta is 1.28.
Worked Example for One-Sample Design
Consider a researcher who wants to determine whether a new culture medium produces a mean bacterial growth rate that differs from the established rate of 10 cells per hour. The researcher expects the new medium to produce a growth rate of 12 cells per hour, and the standard deviation of growth rate measurements is 4 cells per hour.
The effect size delta is 12 minus 5, which equals 7 cells per hour. The standard deviation sigma is 4 cells per hour. The researcher sets alpha at 0.05 and power at 0.80.
Using the formula:
n = (1.96 + 0.84)^2 * 4^2 / 7^2
n = (2.80)^2 * 16 / 49
n = 7.84 * 16 / 49
n = 125.44 / 49
n = 2.56
The calculation produces a sample size of approximately 3. This is a small sample size because the effect size is large relative to the variability. The researcher should verify that the effect size estimate is realistic and that the standard deviation is accurate.
Limitations of One-Sample Calculations
The one-sample formula assumes that the outcome variable is normally distributed and that the standard deviation is known. In practice, the standard deviation is estimated from data, which introduces uncertainty into the calculation. The formula also assumes that the effect size is specified accurately, which is often the most uncertain input.
The one-sample design is less common in biology than the two-sample design because most experiments compare two or more groups. The one-sample design is used when a known reference value exists, such as a standard growth rate, a published mean, or a regulatory threshold.
Two-Sample Designs
Formula for Two Independent Groups
The two-sample design is the most common experimental design in biology research. It compares the means of two independent groups, such as a treatment group and a control group. The sample size formula for a two-sample t-test is:
n per group = 2 (z_alpha/2 + z_beta)^2 sigma^2 / delta^2
In this formula, n per group is the required sample size for each group, z_alpha/2 is the critical value for the significance level, z_beta is the critical value for power, sigma is the standard deviation of the outcome variable, and delta is the expected difference between the two group means.
The factor of 2 in the formula accounts for the fact that two groups are being compared. The total sample size is twice the per-group sample size. This formula assumes equal sample sizes in the two groups, which is the most efficient design when the variability is similar in both groups.
Worked Example: Two-Sample Design
Consider a researcher who wants to compare the effect of two different diets on weight gain in mice. The researcher expects the treatment diet to produce a weight gain of 15 grams and the control diet to produce a weight gain of 10 grams. The standard deviation of weight gain is 5 grams.
The effect size delta is 15 minus 10 = 5 grams. The standard deviation sigma is 5 grams. The researcher sets alpha at 0.05 and power at 0.80.
Using the values 1.96 for z_alpha/2 and 0.84 for z_beta:
n per group = 2 (1.96 + 0.84)^2 5^2 / 5^2
n per group = 2 (2.80)^2 25 / 25
n per group = 2 7.84 1
n per group = 15.68
The sample size per group is approximately 16 mice. The total sample size is 32 mice. The researcher should round up to 16 per group to ensure adequate power.
Unequal Group Sizes
The two-sample formula assumes equal sample sizes in the two groups. In some experiments, unequal group sizes are necessary or desirable. For example, a researcher may have limited access to a rare treatment group or may want to allocate more animals to a control group for ethical reasons.
When group sizes are unequal, the sample size calculation becomes more complex. The total sample size is larger than the equal-size design for the same power. The efficiency of the design is maximized when the groups are equal in size. Researchers should use statistical software to calculate sample sizes for unequal group designs.
Paired Designs
The paired design is a special case of the two-sample design where each observation in one group is matched with an observation in the other group. This design is used when the researcher can control for individual differences by measuring the same subject before and after treatment or by matching subjects on relevant characteristics.
The sample size formula for a paired design is similar to the one-sample formula, but the standard deviation is the standard deviation of the differences between paired observations. The paired design is more efficient than the independent two-sample design when the correlation between paired observations is high, because the variability of the differences is smaller than the variability of the individual observations.
ANOVA Designs
Formula for Multiple Groups
The analysis of variance (ANOVA) design is used when the researcher wants to compare the means of three or more groups. The sample size calculation for ANOVA is more complex than for the two-sample design because it involves multiple comparisons and a different test statistic.
The sample size formula for a one-way ANOVA with k groups is:
n per group = 2 (z_alpha/2 + z_beta)^2 sigma^2 / delta^2
This formula is the same as the two-sample formula, but the interpretation of delta is different. In the ANOVA context, delta is the minimum difference between any two group means that the researcher wants to detect. The formula provides the sample size per group needed to detect this difference with the specified power.
The ANOVA design requires a larger sample size than the two-sample design when the number of groups is large. This is because the test statistic has more degrees of freedom and the critical value for the test is larger. The sample size calculation should account for the number of groups in the design.
Worked Example: ANOVA Design
Consider a researcher who wants to compare the growth rates of bacteria in four different culture media. The researcher expects the minimum difference between any two media to be 3 cells per hour. The standard deviation of growth rate measurements is 4 cells per hour. The researcher sets alpha at 0.05 and power at 0.80.
Using the formula with delta = 3 and sigma = 4:
n per group = 2 (1.96 + 0.84)^2 4^2 / 3^2
n per group = 2 (2.80)^2 16 / 9
n per group = 2 7.84 16 / 9
n per group = 250.88 / 9
n per group = 27.88
The sample size per group is approximately 28. The total sample size is 112 mice. The researcher should round up to 28 per group.
Post Hoc Power Analysis
Post hoc power analysis is the calculation of power after the study has been completed. This analysis uses the observed effect size and variability to calculate the power that the study had to detect the observed effect. Post hoc power is sometimes used to interpret non-significant results, but this practice is controversial.
The main problem with post hoc power analysis is that it uses the observed effect size, which is a biased estimate of the true effect size. When the study is underpowered, the observed effect size is likely to be smaller than the true effect size, which produces a low post hoc power estimate. This low estimate does not provide useful information about the study design.
A more useful approach is to calculate the confidence interval for the effect size and to interpret the non-significant result in the context of the precision of the estimate. The confidence interval shows the range of effect sizes that are consistent with the observed data, which is more informative than a post hoc power calculation.
Proportion-Based Designs
Formula for One-Sample Proportion
Proportion-based designs are used when the outcome variable is binary, such as the presence or absence of a trait, the success or failure of a treatment, or the proportion of cells that respond to a stimulus. The sample size formula for a one-sample proportion test is:
n = (z_alpha/2 sqrt(p0(1-p0)) + z_beta sqrt(p1(1-p1)))^2 / (p1-p0)^2
In this formula, p0 is the proportion under the null hypothesis, p1 is the expected proportion under the alternative hypothesis, z_alpha/2 is the critical value for the significance level, and z_beta is the critical value for the power.
The formula accounts for the variability of the proportion under both the null and alternative hypotheses. The variability of a proportion is highest when the proportion is 0.5 and decreases as the proportion approaches 0 or 1.
Worked Example: One-Sample Proportion
Consider a researcher who wants to determine whether the proportion of cells that survive a particular treatment is different from the known survival rate of 0.50. The researcher expects the survival rate to be 0.60. The researcher sets alpha at 0.05 and power at 0.80.
Using the values p0 = 0.50, p1 = 0.60, z_alpha/2 = 1.96, and z_beta = 0.84:
n = (1.96 sqrt(0.50 0.50) + 0.84 sqrt(0.60 0.40))^2 / (0.60 - 0.50)^2
n = (1.96 0.50 + 0.84 0.49)^2 / 0.01
n = (0.98 + 0.41)^2 / 0.01
n = (1.39)^2 / 0.01
n = 1.93 / 0.01
n = 193
The sample size is approximately 193 observations. This is a large sample size because the expected difference between the null and alternative proportions is small.
Formula for Two-Sample Proportion
The two-sample proportion design compares the proportions of two independent groups. The sample size formula for this design is:
n per group = (z_alpha/2 sqrt(2 p_bar(1-p_bar)) + z_beta * sqrt(p1(1-p1) + p2(1-p2)))^2 / (p1 - p2)^2
In this formula, p1 and p2 are the expected proportions in the two groups, p_bar is the average of p1 and p2, z_alpha/2 is the critical value for the significance level, and z_beta is the critical value for the power.
The formula accounts for the variability of the proportions in both groups. The term p_bar is used for the null hypothesis, which assumes that the two proportions are equal. The term p1(1-p1) + p2(1-p2) is used for the alternative hypothesis, which allows the proportions to differ.
Worked Example: Two-Sample Proportion
Consider a researcher who wants to compare the proportion of cells that express a certain protein in two different cell lines. The researcher expects the proportion to be 0.30 in the first cell line and 0.50 in the second cell line. The researcher sets alpha at 0.05 and power at 0.80.
Using the formula with p1 = 0.30, p2 = 0.50, p_bar = 0.40, z_alpha/2 = 1.96, and z_beta = 0.84:
n per group = (1.96 sqrt(2 0.40 0.60) + 0.84 sqrt(0.30 0.70 + 0.50 0.50))^2 / (0.50 - 0.30)^2
n per group = (1.96 sqrt(0.48) + 0.84 sqrt(0.21 + 0.25))^2 / (0.20)^2
n per group = (1.96 0.69 + 0.84 sqrt(0.46))^2 / 0.04
n per group = (1.35 + 0.84 * 0.68)^2 / 0.04
n per group = (1.35 + 0.57)^2 / 0.04
n per group = (1.92)^2 / 0.04
n per group = 3.69 / 0.04
n per group = 92.25
The sample size per group is approximately 93. The total sample size is 186 observations.
Practical Implementation Steps
Step 1: Define the Research Question and Design
The first step in sample size calculation is to define the research question and the experimental design. The researcher must specify the primary outcome variable, the comparison groups, and the statistical test that will be used for the analysis. This step is critical because the sample size formula depends on the design.
The research question should be specific and answerable. For example, the question should specify the outcome variable, the comparison groups, and the direction of the expected effect. A vague research question produces a vague sample size calculation.
The experimental design should be determined by the research question and the practical constraints of the study. The design should be chosen to minimize bias and maximize the efficiency of the study. The sample size calculation should be based on the primary analysis, not on secondary analyses.
Step 2: Specify the Inputs
The researcher must specify the inputs for the sample size calculation. These inputs include the significance level, the power, the effect size, and the standard deviation. The significance level and power are typically set at conventional values, but the effect size and standard deviation require estimation.
The effect size should be based on the smallest effect that is biologically or clinically meaningful. This is the effect that the researcher wants to be able to detect. The effect size should not be based on the effect that the researcher expects to observe, because this can lead to an underpowered study.
The standard deviation should be based on prior data from the same laboratory, published literature, or a pilot study. The standard deviation should be estimated conservatively to avoid an underpowered study. If the standard deviation is uncertain, the researcher should consider using a range of values in the calculation.
Step 3: Calculate the Sample Size
The researcher should calculate the sample size using the appropriate formula for the design. The calculation can be done by hand using the formulas in this article, or using statistical software. The software should be validated against the formula to ensure that the calculation is correct.
The sample size should be rounded up to the nearest integer. The researcher should also consider the expected dropout rate or the number of observations that will be lost due to technical failures. The final sample size should be increased to account for these losses.
Step 4: Perform a Sensitivity Analysis
A sensitivity analysis is a series of sample size calculations using different values for the inputs. This analysis shows how the sample size changes with the effect size, the standard deviation, and the power. The sensitivity analysis helps the researcher understand the robustness of the sample size calculation.
The sensitivity analysis should include a range of effect sizes and standard deviations. The researcher should identify the sample size that is adequate for the most conservative scenario. This approach ensures that the study will have adequate power even if the inputs are not accurate.
Step 5: Document the Calculation
The sample size calculation should be documented in the study protocol or the research plan. The documentation should include the formula used, the inputs, and the resulting sample size. This documentation is important for transparency and for the review of the study.
The documentation should also include the rationale for the choice of inputs. This includes the source of the effect size estimate and the standard deviation estimate. The documentation should be detailed enough that another researcher can reproduce the calculation.
Records and Measurements
Data Management and Sharing
The National Institutes of Health has established expectations for data management and sharing in research. The NIH Data Management and Sharing Policy requires researchers to plan for the management and sharing of data generated by NIH-funded research. This policy applies to the data that will be collected in the study, including the sample size and the data that will be used for the analysis.
The data management plan should describe how the data will be collected, stored, and shared. The plan should also describe the sample size calculation and the assumptions that were used. The plan should be included in the grant application and updated as the study progresses.
The data management plan is important for the reproducibility of the research. The plan should ensure that the data are available for other researchers to verify the results and to conduct secondary analyses. The plan should also ensure that the data are stored securely and that the privacy of the participants is protected.
Documentation of the Sample Size Calculation
The sample size calculation should be documented in the research protocol and in the final research report. The documentation should include the formula used, the inputs, and the resulting sample size. The documentation should also include the sensitivity analysis and the rationale for the choice of inputs.
The documentation is important for the review of the research. The reviewers will want to know whether the sample size is adequate for the research question. The documentation should be clear and complete enough that the reviewers can evaluate the adequacy of the sample size.
Reporting Guidelines
The EQUATOR Network provides a collection of reporting guidelines for health research. These guidelines are designed to improve the quality and transparency of research reporting. The guidelines include recommendations for reporting the sample size calculation in research reports.
The reporting guidelines recommend that the sample size calculation be described in the methods section of the research report. The description should include the formula used, the inputs, and the resulting sample size. The description should also include the rationale for the choice of inputs.
The reporting guidelines are important for the reproducibility of the research. The guidelines ensure that the sample size calculation is transparent and that the reader can evaluate the adequacy of the sample size. The guidelines also ensure that the sample size calculation is consistent with the research question and the design.
Common Failure Patterns
Underpowered Studies
The most common failure pattern in sample size calculation is an underpowered study. This occurs when the sample size is too small to detect the effect of interest. An underpowered study is a waste of resources because it is unlikely to produce a statistically significant result, even if the effect is real.
Underpowered studies can occur for several reasons. The researcher may have used an effect size that is too large, a standard deviation that is too small, or a power that is too low. The researcher may also have failed to account for the dropout rate or the number of missing data.
The consequence of an underpowered study is a high probability of a false negative result. The study may fail to detect a real effect, which can lead to a wrong conclusion. The study may also produce an imprecise estimate of the effect size, which is not useful for decision-making.
Overpowered Studies
An overpowered study is a study with a sample size that is larger than necessary to detect the effect of interest. This can occur when the researcher uses a conservative effect size or a high power. An overpowered study is a waste of resources, but it is less problematic than an underpowered study.
The consequence of an overpowered study is that the study may detect a small effect that is not biologically meaningful. The study may also be more expensive and time-consuming than necessary. The researcher should balance the need for power against the cost of collecting more data.
Incorrect Inputs
The sample size calculation is only as good as the inputs. If the effect size or the standard deviation is incorrect, the sample size will be incorrect. The researcher should use the best available estimates for the inputs and should perform a sensitivity analysis to assess the robustness of the sample size.
The effect size is the most difficult input to estimate. The researcher should use the smallest effect that is biologically meaningful, not the effect that is expected. The standard deviation should be estimated from prior data or a pilot study. The researcher should be conservative in the estimates to avoid an underpowered study.
Failure to Account for Dropout
The sample size calculation should account for the expected dropout rate or the number of missing data. The researcher should increase the sample size to account for the observations that will be lost. The dropout rate should be estimated from prior studies or from the design of the study.
The failure to account for dropout can result in an underpowered study. The researcher should calculate the sample size for the number of observations that will be available for the analysis, not the number of observations that will be collected. The sample size should be increased by the expected dropout rate.
Limitations and Considerations
Assumptions of the Formulas
The sample size formulas in this article are based on several assumptions. The formulas assume that the outcome variable is normally distributed, that the standard deviation is known, and that the observations are independent. These assumptions are not always met in practice.
The normality assumption is important for the t-test and ANOVA. If the outcome variable is not normally distributed, the sample size calculation may be inaccurate. The researcher should consider using a nonparametric test or a transformation of the data.
The assumption of a known standard deviation is also important. In practice, the standard deviation is estimated from data, which introduces uncertainty. The researcher should use a conservative estimate of the standard deviation to ensure adequate power.
Ethical Considerations
The sample size calculation has ethical implications. A study that is underpowered is unethical because it exposes subjects to the risks of the research without the potential benefit of a meaningful result. A study that is overpowered is unethical because it exposes more subjects than necessary to the risks of the research.
The researcher should use the sample size calculation to minimize the number of subjects while maintaining adequate power. The researcher should also consider the welfare of the subjects and the ethical principles of the research. The sample size calculation is an important part of the ethical design of the research.
Professional Escalation
The researcher should consult a statistician or a biostatistician when the sample size calculation is complex or when the researcher is uncertain about the inputs. The statistician can help the researcher choose the appropriate formula and estimate the inputs. The statistician can also help the researcher interpret the results of the sample size calculation.
The researcher should also consult the relevant guidelines and regulations for the research. The guidelines may require a specific sample size or a specific approach to the sample size calculation. The researcher should be aware of the requirements and should follow them.
Frequently Asked Questions
What is the difference between sample size and power?
Sample size is the number of observations in the study. Power is the probability that the study will detect a true effect. The sample size is a key determinant of power. A larger sample size increases the power of the study, but the relationship is not linear.
How do I choose the effect size for the sample size calculation?
The effect size should be the smallest effect that is biologically or clinically meaningful. This is the effect that the researcher wants to be able to detect. The effect size should not be based on the effect that is expected to be observed, because this can lead to an underpowered study.
What is the difference between a one-sample and a two-sample design?
A one-sample design compares the mean of a single group to a known value. A two-sample design compares the means of two independent groups. The sample size formula for the two-sample design is different from the one-sample design because the two-sample design has more variability.
How do I account for multiple comparisons in the sample size calculation?
Multiple comparisons increase the risk of a Type I error. The researcher should adjust the alpha level for the multiple comparisons, which will increase the sample size. The Bonferroni correction is a common method for adjusting the alpha level.
What is the difference between a paired and an unpaired design?
A paired design matches each observation in one group with an observation in the other group. An unpaired design does not match the observations. The paired design is more efficient than the unpaired design when the correlation between the observations is high.
How do I calculate the sample size for a proportion-based outcome?
The sample size for a proportion-based outcome is calculated using the formula for the proportion test. The formula requires the expected proportions under the null and alternative hypotheses. The sample size is larger when the difference between the proportions is small.
What is the role of the standard deviation in the sample size calculation?
The standard deviation is a measure of the variability of the outcome variable. A larger standard deviation requires a larger sample size to achieve the same power. The standard deviation should be estimated from prior data or a pilot study.
What should I do if the sample size calculation produces a very large sample size?
A very large sample size may indicate that the effect size is too small or that the standard deviation is too large. The researcher should consider whether the effect size is biologically meaningful and whether the standard deviation can be reduced. The researcher may also consider a different design or a different outcome variable.
Using the Evidence
| Source | Best use in this topic | Important limitation |
|---|---|---|
| Research Methods Resources | official guidance | Check the linked page for current local requirements |
| EQUATOR Network | official guidance | Check the linked page for current local requirements |
| Core Practices | official guidance | Check the linked page for current local requirements |
Related Diagnostic Guides
- SNAP Tests versus Reference Lab ELISA for Feline Retrovirus Screening
- Nanotechnology in Rapid Viral Diagnostic Tests
- Animal Biology
References and Further Reading
- Research Methods Resources. National Library of Medicine.
- EQUATOR Network. EQUATOR Network.
- Core Practices. Committee on Publication Ethics.
- NIH Grants and Funding. National Institutes of Health.
- ORCID for Researchers. ORCID.
- Data Management and Sharing Policy. National Institutes of Health.
- Laboratory Biosafety Manual. World Health Organization.
- NCBI Literature Resources. National Center for Biotechnology Information.
- Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range.. BMC medical research methodology, 2014.
- Sample Size Estimation in Clinical Research: From Randomized Controlled Trials to Observational Studies.. Chest, 2020.
- Sample size estimation in diagnostic test studies of biomedical informatics.. Journal of biomedical informatics, 2014.
This article is educational and does not replace validated laboratory procedures, institutional biosafety review, manufacturer instructions, or professional interpretation.