How to Choose Effect Size for Sample Size Calculation

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-27

Quick Answer

Select effect size from pilot data, published literature, or a minimally important difference, then verify it against biological plausibility before calculating sample size.
Report the effect size, its source, and the uncertainty around it in the study protocol and final manuscript so reviewers can assess the power analysis.
Effect size is a standardized or raw measure of the expected biological signal, and a poorly chosen value can make a study underpowered or impractically large.

At a Glance

Effect Size Source	Typical Use	Key Limitation	Documentation Needed
Pilot data	First estimate of variability and mean difference	Small pilot samples produce imprecise estimates	Pilot sample size, raw data, and confidence intervals
Published literature	Pre-specified hypothesis with prior evidence	Publication bias may inflate reported effects	Full citation and effect size extraction method
Minimally important difference	Clinical or practical relevance threshold	Requires stakeholder agreement on what matters	Justification for the threshold and its origin
Expert consensus	When no data exist	Subjective and difficult to audit	Written rationale and named contributors

The Role of Effect Size in Sample Size Calculation

Sample size calculation requires four inputs: the significance level, the statistical power, the variability of the outcome, and the effect size. The effect size is the magnitude of the biological or clinical difference the study is designed to detect. It is the input that researchers most often choose arbitrarily, and it is the input that most strongly determines the final sample size.

A small effect size produces a large sample size requirement. A large effect size produces a smaller sample size requirement but risks missing a biologically meaningful difference that is smaller than the assumed value. The researcher must therefore select an effect size that is both biologically plausible and practically meaningful for the study question.

The effect size can be expressed in raw units, such as a difference in mean blood glucose concentration, or in standardized units, such as Cohen d or eta squared. Standardized effect sizes allow comparison across studies that use different measurement scales. Raw effect sizes are easier to interpret in the context of the specific experiment.

The choice of effect size is not a purely statistical decision. It depends on the research question, the prior evidence, the feasibility of recruiting or breeding the required number of subjects, and the consequences of missing a true effect. The researcher must document the source of the effect size and the reasoning behind the choice.

Core Principles for Selecting an Effect Size

The Relationship Between Effect Size and Sample Size

The sample size required for a study increases as the effect size decreases. This relationship is nonlinear. Halving the effect size can quadruple the required sample size, depending on the statistical test and the variance structure. Researchers who underestimate the effect size may design a study that is too small to detect the difference they care about. Researchers who overestimate the effect size may design a study that is too large to complete within budget or time constraints.

The power of a study is the probability of detecting a true effect of a given size. Power is typically set at 0.80 or 0.90, meaning the study has an 80 percent or 90 percent chance of detecting the effect if it exists. The effect size is the magnitude of the difference the study is powered to detect. The sample size is the number of independent experimental units needed to achieve that power.

The relationship between these quantities is deterministic once the statistical test and the variability of the outcome are specified. The researcher does not choose the sample size directly. The researcher chooses the effect size, the power, the significance level, and the variability, and the sample size follows from those choices.

Standardized Versus Raw Effect Sizes

Standardized effect sizes express the difference in units of the standard deviation. Cohen's d is the difference between two group means divided by the pooled standard deviation. A d of 0.2 is considered small, 0.5 is considered medium, and 0.8 is considered large in the social sciences. These thresholds are not universal and may not apply to biological experiments where the outcome is tightly controlled.

Raw effect sizes express the difference in the original units of the measurement. A raw effect size of 5 milligrams per deciliter is easier to interpret for a clinician or a farmer than a standardized effect size. Raw effect sizes are preferred when the measurement unit has a clear biological meaning and when the variability is well understood.

The choice between standardized and raw effect sizes depends on the audience and the purpose of the study. Standardized effect sizes are useful for meta-analysis and for comparing studies that use different measurement scales. Raw effect sizes are useful for designing a study with a specific biological threshold in mind.

The Role of Variability in Effect Size Selection

The effect size is not the only determinant of sample size. The variability of the outcome, expressed as the standard deviation or the coefficient of variation, also affects the sample size. A highly variable outcome requires a larger sample size to detect the same difference. A homogeneous outcome requires a smaller sample size.

The researcher must estimate the variability from pilot data, from the literature, or from prior experiments in the same laboratory. The variability estimate is often as uncertain as the effect size estimate. The researcher should report the variability estimate and its source in the power analysis.

The coefficient of variation is a useful measure when the standard deviation increases with the mean. This is common in biological measurements such as enzyme activity, hormone concentration, and cell count. The coefficient of variation is the standard deviation divided by the mean. It is expressed as a percentage and is useful for comparing variability across groups with different means.

Sources of Effect Size Estimates

Pilot Data

Pilot data are the most direct source of effect size estimates. A pilot study is a small-scale version of the planned experiment. It provides an estimate of the mean difference between groups and the variability of the outcome. The pilot data can be used to calculate the effect size and the required sample size for the full study.

The limitation of pilot data is the small sample size. A pilot study with 10 subjects per group produces a wide confidence interval around the effect size. The effect size estimate from a pilot study may be imprecise and may not reflect the true effect in the population. The researcher should report the pilot sample size and the confidence interval for the effect size.

The pilot study should use the same measurement procedures, the same experimental conditions, and the same subject population as the full study. A pilot study that differs in any of these aspects may produce an effect size estimate that does not apply to the full study. The pilot study should also be analyzed with the same statistical test planned for the full study.

Published Literature

Published literature is a common source of effect size estimates. A meta-analysis provides a pooled effect size estimate across multiple studies. The pooled estimate is more precise than the estimate from any single study. The researcher should use the pooled effect size from a meta-analysis when one is available.

A single published study can also provide an effect size estimate. The effect size from a single study is less precise than the pooled estimate from a meta-analysis. The researcher should consider the quality of the study, the similarity of the study population to the planned study population, and the measurement methods used in the study.

The researcher should be aware that published effect sizes may be inflated by publication bias. Studies with large effects are more likely to be published than studies with small effects. The effect size from the literature may therefore be larger than the true effect size. The researcher should consider this bias when selecting an effect size from the literature.

Minimally Important Difference

The minimally important difference is the smallest difference that is considered biologically or clinically meaningful. This value is determined by the research question and the stakeholders. It is not derived from the data. It is a judgment about the smallest effect that matters.

The minimally important difference is often used in clinical trials and in regulatory studies. It is the difference that the study is designed to detect. The sample size is calculated to detect this difference with the specified power and significance level.

The minimally important difference should be justified in the study protocol. The justification should explain why this particular difference is meaningful. The justification may be based on clinical experience, on the results of a survey of stakeholders, or on the relationship between the outcome and a downstream consequence.

Field Consensus and Expert Judgment

When no pilot data and no published literature are available, the researcher may rely on expert judgment. This is the least reliable source of effect size estimates. The judgment is subjective and may not reflect the true effect size. The researcher should document the basis for the judgment and the names of the experts who provided it.

Expert judgment can be formalized using a Delphi process or a structured elicitation method. These methods involve multiple experts who provide independent estimates and then discuss and revise their estimates. The result is a consensus estimate that is more reliable than the estimate from a single expert.

The researcher should treat expert judgment as a last resort. The effect size estimate from expert judgment should be tested in a pilot study before the full study is designed. The pilot study will provide an empirical estimate of the effect size that can be compared with the expert judgment.

Practical Workflow for Effect Size Selection

Step 1: Define the Primary Outcome

The primary outcome is the measurement that will be used to answer the main research question. The primary outcome must be defined before the effect size is selected. The primary outcome should be a single measurement, not a composite of multiple measurements. The primary outcome should be measured in the same way for all subjects in the study.

The primary outcome should be relevant to the research question and should be responsive to the intervention or the condition being studied. The primary outcome should have a known distribution and a known variability in the study population. The researcher should consult the literature or pilot data to determine the distribution and the variability of the primary outcome.

Step 2: Identify the Effect Size Source

The researcher should identify the source of the effect size estimate before calculating the sample size. The source should be pilot data, published literature, a minimally important difference, or expert consensus. The source should be documented in the study protocol.

The researcher should prefer the most reliable source that is available. Pilot data from a well-designed pilot study are more reliable than expert judgment. A meta-analysis is more reliable than a single study. The minimally important difference is the most appropriate source when the study is designed to detect a specific clinical or regulatory threshold.

Step 3: Estimate the Variability

The variability of the primary outcome must be estimated. The variability is expressed as the standard deviation or the coefficient of variation. The variability should be estimated from the same source as the effect size. If the effect size is from a meta-analysis, the variability should be the pooled standard deviation from the meta-analysis. If the effect size is from a pilot study, the variability should be the standard deviation from the pilot study.

The variability should be estimated for the control group and for the treatment group. The variability may differ between groups. The researcher should use the variability that is most appropriate for the planned statistical test. For a two-sample t-test, the pooled standard deviation is used. For a paired t-test, the standard deviation of the differences is used.

Step 4: Calculate the Sample Size

The sample size is calculated using the effect size, the variability, the significance level, and the power. The calculation can be performed using a statistical software package or an online calculator. The researcher should specify the statistical test, the significance level, the power, and the effect size.

The sample size calculation should be performed for the primary outcome. The sample size for secondary outcomes may be different. The researcher should report the sample size for the primary outcome and the assumptions used in the calculation.

Step 5: Perform a Sensitivity Analysis

The sensitivity analysis examines how the sample size changes when the effect size or the variability changes. The researcher should calculate the sample size for a range of effect sizes and a range of variability estimates. The sensitivity analysis shows how robust the sample size is to the assumptions.

The sensitivity analysis should include the lower and upper bounds of the confidence interval for the effect size. The sample size should be calculated for the lower bound, the point estimate, and the upper bound. The researcher should report the range of sample sizes and the assumptions that produce each sample size.

Step 6: Document the Effect Size and Its Source

The effect size and its source should be documented in the study protocol and in the final manuscript. The documentation should include the effect size value, the source of the effect size, the variability estimate, and the sample size calculation. The documentation should be sufficient for a reviewer to reproduce the sample size calculation.

The documentation should also include the sensitivity analysis. The sensitivity analysis shows the reviewer that the sample size is robust to the assumptions. The documentation should be included in the methods section of the manuscript or in a supplementary file.

Common Failure Patterns in Effect Size Selection

Using the Effect Size from a Previous Study Without Adjustment

A common failure is to use the effect size from a previous study without considering the differences between the previous study and the planned study. The previous study may have used a different population, a different measurement method, or a different experimental condition. The effect size from the previous study may not apply to the planned study.

The researcher should compare the planned study to the previous study. The comparison should include the population, the measurement method, the experimental condition, and the outcome definition. If the studies differ in any of these aspects, the effect size should be adjusted or a different source should be used.

Using a Standardized Effect Size Without Context

A common failure is to use a standardized effect size, such as Cohen's d, without considering the context. The standardized effect size is a unitless measure that depends on the variability of the outcome. A d of 0.5 may be a large effect in one population and a small effect in another population.

The researcher should interpret the standardized effect size in the context of the outcome and the population. The researcher should also report the raw effect size and the variability. The raw effect size is more interpretable than the standardized effect size.

Ignoring the Uncertainty in the Effect Size

A common failure is to treat the effect size as a fixed value. The effect size is an estimate and has uncertainty. The uncertainty is expressed as a confidence interval. The researcher should consider the confidence interval when selecting the effect size and when calculating the sample size.

The researcher should calculate the sample size for the lower bound and the upper bound of the confidence interval. The sample size for the lower bound is the largest sample size. The sample size for the upper bound is the smallest sample size. The researcher should choose a sample size that is adequate for the lower bound of the effect size.

Using a Small Pilot Study to Estimate the Effect Size

A common failure is to use a small pilot study to estimate the effect size. A pilot study with a small sample size produces an imprecise effect size estimate. The imprecision is expressed as a wide confidence interval. The effect size estimate from a small pilot study may be far from the true effect size.

The researcher should use the confidence interval from the pilot study to calculate the sample size. The sample size should be calculated for the lower bound of the confidence interval. The researcher should also consider the possibility that the pilot study overestimates the effect size.

Confusing the Effect Size with the Significance Level

A common failure is to confuse the effect size with the significance level. The significance level is the probability of a false positive. The effect size is the magnitude of the difference. The significance level is typically set at 0.05. The effect size is the difference that the study is designed to detect.

The researcher should not adjust the effect size to achieve a desired sample size. The effect size should be based on the biological question and the prior evidence. The sample size should be calculated from the effect size, not the other way around.

Observations and Measurements for Effect Size Estimation

Recording the Variability of the Outcome

The variability of the outcome should be recorded in the pilot study and in the full study. The variability is expressed as the standard deviation or the coefficient of variation. The variability should be recorded for each group in the study. The variability should be recorded in the study protocol and in the final manuscript.

The variability of the outcome can change during the study. The variability may increase if the measurement method is not consistent or if the experimental conditions change. The researcher should monitor the variability during the study and should report any changes.

Recording the Effect Size in the Study Protocol

The effect size should be recorded in the study protocol before the study begins. The effect size should be recorded with its source and its justification. The effect size should be recorded in the same units as the outcome. The effect size should be recorded in the protocol and in the final manuscript.

The effect size should be recorded in the statistical analysis plan. The statistical analysis plan should describe the statistical test, the significance level, the power, and the effect size. The statistical analysis plan should be written before the data are collected.

Recording the Sample Size Calculation

The sample size calculation should be recorded in the study protocol. The calculation should include the effect size, the variability, the significance level, and the power. The calculation should include the software or the formula used. The calculation should be reproducible by a reviewer.

The sample size calculation should be recorded in the final manuscript. The calculation should be described in the methods section. The calculation should include the effect size, the variability, the significance level, and the power. The calculation should include the source of the effect size and the variability.

Quality Controls for Effect Size Selection

Independent Review of the Effect Size

The effect size should be reviewed by an independent researcher. The independent reviewer should assess the source of the effect size and the justification. The independent reviewer should assess whether the effect size is biologically plausible and statistically appropriate. The independent reviewer should provide a written assessment.

The independent review should be conducted before the study begins. The review should be documented in the study protocol. The review should be recorded in the final manuscript. The independent review is a quality control measure that reduces the risk of a biased effect size.

Pre-Specification of the Effect Size

The effect size should be pre-specified in the study protocol. The pre-specification prevents the researcher from changing the effect size after the data are collected. The pre-specification also prevents the researcher from selecting the effect size that produces a desired sample size.

The pre-specification should include the effect size value, the source of the effect size, and the justification. The pre-specification should be recorded in the study protocol and in the final manuscript. The pre-specification should be reviewed by an independent researcher.

Sensitivity Analysis as a Quality Control

The sensitivity analysis is a quality control for the effect size. The sensitivity analysis examines how the sample size varies with the effect size and the variability. The sensitivity analysis should be performed before the study begins. The sensitivity analysis should be recorded in the study protocol and in the final manuscript.

The sensitivity analysis should include the lower and upper bounds of the confidence interval for the effect size. The sample size should be calculated for the lower bound and the upper bound. The sensitivity analysis should be reported in the final manuscript.

Limitations of Effect Size Estimates

The Effect Size Is an Estimate

The effect size is an estimate, not a known value. The effect size is estimated from pilot data, from the literature, or from expert judgment. The estimate has uncertainty. The uncertainty is expressed as a confidence interval. The researcher should report the confidence interval for the effect size.

The effect size estimate may be biased. The bias may be due to publication bias, to the small sample size of the pilot study, or to the subjective nature of expert judgment. The researcher should consider the potential bias in the effect size estimate.

The Effect Size May Not Be Generalizable

The effect size may not be generalizable to other populations or to other experimental conditions. The effect size is specific to the population and the conditions of the study. The effect size may be different in a different population or under different conditions.

The researcher should consider the generalizability of the effect size. The researcher should compare the planned study to the study that produced the effect size. The researcher should consider the differences in the population, the measurement method, and the experimental conditions.

The Effect Size May Change During the Study

The effect size may change during the study. The effect size may be different in the full study than in the pilot study. The effect size may be different in the early part of the study than in the later part of the study. The researcher should monitor the effect size during the study.

The researcher should not change the sample size during the study based on the observed effect size. The sample size should be determined before the study begins. The sample size should be based on the pre-specified effect size and the pre-specified variability.

Reporting the Effect Size in the Manuscript

The effect size should be reported in the methods section of the manuscript. The report should include the effect size value, the source of the effect size, and the justification. The report should include the variability estimate and the sample size calculation. The report should include the sensitivity analysis.

The effect size should be reported in the results section of the manuscript. The report should include the observed effect size and the confidence interval. The observed effect size should be compared to the pre-specified effect size. The comparison should be discussed in the discussion section.

The effect size should be reported in the abstract of the manuscript. The abstract should include the pre-specified effect size and the sample size. The abstract should include the observed effect size and the confidence interval. The abstract should be concise and should not include the full details of the effect size selection.

Reporting Guidelines and Transparency

The reporting guidelines for the study should be selected before the study begins. The EQUATOR Network provides a searchable database of reporting guidelines for different study types. The reporting guidelines should be followed in the manuscript. The reporting guidelines should be cited in the methods section.

The reporting guidelines for a randomized controlled trial include the CONSORT statement. The reporting guidelines for an observational study include the STROBE statement. The reporting guidelines for a diagnostic accuracy study include the STARD statement. The reporting guidelines for a systematic review include the PRISMA statement.

The reporting guidelines should be followed for the effect size and the sample size calculation. The guidelines should be followed for the reporting of the observed effect size and the confidence interval. The guidelines should be followed for the reporting of the sensitivity analysis.

The transparency of the effect size selection is important for the reproducibility of the study. The effect size and the sample size calculation should be reported in sufficient detail for a reviewer to reproduce the calculation. The effect size and the sample size calculation should be reported in the methods section or in a supplementary file.

Data Management and Sharing

The data from the pilot study and the full study should be managed according to the data management and sharing policy. The data should be described in a data management and sharing plan. The plan should describe the data that will be collected, the data that will be shared, and the data that will be preserved.

The data management and sharing plan should be submitted with the grant application. The plan should be reviewed by the funding agency. The plan should be updated during the study if the data collection or the data sharing changes.

The data from the pilot study should be shared with the full study. The data from the pilot study should be shared in a data repository. The data from the full study should be shared in a data repository. The data should be shared in a format that is usable by other researchers.

The data sharing should be described in the manuscript. The manuscript should describe the data that are shared and the repository where the data are shared. The manuscript should describe the data sharing in the methods section or in a data availability statement.

Researcher Identity and Documentation

The researcher should maintain a record of their research activities. The record should include the study protocol, the data, the analysis, and the manuscript. The record should be maintained in a secure location. The record should be maintained for the duration of the study and for a period after the study.

The researcher should use a persistent identifier for their research. The persistent identifier should be used in the manuscript and in the data repository. The persistent identifier should be used in the grant application and in the data management plan.

The researcher should maintain their research record. The research record should include the researcher's publications, the researcher's data, and the researcher's peer review activities. The research record should be updated regularly.

The researcher should use the persistent identifier in the manuscript. The persistent identifier should be included in the manuscript in the author information section. The persistent identifier should be included in the data repository and in the grant application.

Grant Application and Review

The effect size and the sample size calculation should be described in the grant application. The grant application should describe the effect size, the source of the effect size, and the justification. The grant application should describe the sample size calculation and the sensitivity analysis.

The grant application should be reviewed by the funding agency. The review should assess the effect size and the sample size calculation. The review should assess the feasibility of the study. The review should assess the scientific merit of the study.

The grant application should be revised based on the review. The revision should address the comments of the reviewers. The revision should be submitted to the funding agency. The revision should be reviewed again by the funding agency.

The grant application should be managed by the researcher. The researcher should track the grant application and the review. The researcher should track the award and the management of the award. The researcher should track the reporting requirements of the award.

Publication Ethics

The effect size and the sample size calculation should be reported honestly. The researcher should not misrepresent the effect size or the sample size calculation. The researcher should not misrepresent the source of the effect size or the justification.

The researcher should follow the core practices of publication ethics. The core practices include authorship, peer review, data, conflicts, misconduct, and publication ethics. The researcher should follow the core practices in the manuscript and in the publication process.

The researcher should disclose any conflicts of interest. The conflicts of interest should be disclosed in the manuscript and in the grant application. The conflicts of interest should be disclosed to the journal and to the funding agency.

The researcher should not engage in research misconduct. The research misconduct includes fabrication, falsification, and plagiarism. The research misconduct should be reported to the institution and to the funding agency.

Professional Escalation Criteria

The researcher should escalate the effect size selection to a supervisor or a statistician in the following situations:

The effect size is uncertain and the researcher cannot identify a reliable source.
The effect size produces a sample size that is not feasible within the budget or the time constraints.
The effect size is inconsistent with the prior evidence or the biological plausibility.
The effect size is challenged by a reviewer or a collaborator.
The effect size is changed after the study begins.

The researcher should escalate the sample size calculation to a statistician in the following situations:

The statistical test is complex and the researcher is not confident in the calculation.
The variability is not well understood and the researcher is not confident in the estimate.
The sample size calculation is challenged by a reviewer or a collaborator.
The sample size calculation is changed after the study begins.

The researcher should escalate the data management to the institution in the following situations:

The data management plan is not compliant with the funding agency policy.
The data sharing is not compliant with the funding agency policy.
The data are sensitive and the researcher is not confident in the data sharing plan.

Frequently Asked Questions

What is the most reliable source for an effect size estimate?

Pilot data from a well-designed pilot study are the most reliable source when the pilot study uses the same population, measurement method, and experimental conditions as the full study. A meta-analysis from the published literature is the most reliable source when the meta-analysis is current and the studies are similar to the planned study. The minimally important difference is the most appropriate source when the study is designed to detect a clinical or regulatory threshold.

How do I choose between a standardized and a raw effect size?

Use a raw effect size when the measurement units have a clear biological meaning and when the variability is well understood. Use a standardized effect size when you need to compare the effect across studies that use different measurement scales. Report both the raw and the standardized effect size in the manuscript.

What is the minimally important difference and how is it determined?

The minimally important difference is the smallest difference in the outcome that is considered biologically or clinically meaningful. It is determined by the research question and the stakeholders, not by the data. It may be based on the literature, on a survey of stakeholders, or on a relationship between the outcome and a clinical consequence.

How does the variability of the outcome affect the sample size?

The sample size increases as the variability of the outcome increases. The variability is expressed as the standard deviation or the coefficient of variation. The variability should be estimated from the same source as the effect size. The variability should be reported in the study protocol and in the final manuscript.

What is a sensitivity analysis and why is it important?

A sensitivity analysis examines how the sample size varies with the effect size and the variability. The sensitivity analysis is important because the effect size and the variability are estimates with uncertainty. The sensitivity analysis shows the range of sample sizes and the assumptions that produce them. The sensitivity analysis should be reported in the study protocol and in the final manuscript.

What should be reported in the manuscript about the effect size?

The manuscript should report the pre-specified effect size, the source of the effect size, and the justification. The manuscript should report the variability estimate and the sample size calculation. The manuscript should report the sensitivity analysis. The manuscript should report the observed effect size and the confidence interval.

What is the role of reporting guidelines in effect size reporting?

Reporting guidelines provide a checklist of items that should be reported in a manuscript. The guidelines are specific to the study type. The guidelines should be used to ensure that the effect size and the sample size calculation are reported in sufficient detail. The guidelines should be cited in the manuscript.

When should I escalate the effect size selection to a statistician?

You should escalate the effect size selection to a statistician when the effect size is not well understood, when the effect size produces an infeasible sample size, when the effect size is inconsistent with the prior evidence, or when the effect size is challenged by a reviewer or a collaborator. You should also escalate when the statistical test is complex and you are not confident in the calculation.

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

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.
Sample size determination and power analysis using the G*Power software.. Journal of educational evaluation for health professions, 2021.
Sample size, power and effect size revisited: simplified and practical approaches in pre-clinical, clinical and laboratory studies.. Biochemia medica, 2021.
Estimating the sample size for a pilot randomised trial to minimise the overall trial sample size for the external pilot and main trial for a continuous outcome variable.. Statistical methods in medical research, 2016.

This article is educational and does not replace validated laboratory procedures, institutional biosafety review, manufacturer instructions, or professional interpretation.