# Estimating Standard Deviation for Sample Size
## Quick Answer

- Estimate the standard deviation for sample size calculations from pilot data, published literature, or expert opinion when prior data are scarce.
- Use the resource equation approach when you cannot assume a standard deviation or effect size, particularly in animal studies.
- Standard deviation estimates from small pilot samples carry substantial uncertainty and should be treated as provisional values, not fixed constants.

## At a Glance

| Estimation Method | Data Required | Best Use Case | Key Limitation |
| --- | --- | --- | --- |
| Pilot study standard deviation | 10 to 20 observations from your own system | When your experimental conditions differ from published work | Small pilot samples produce imprecise variance estimates |
| Published literature values | Reported mean, standard deviation, median, range, or interquartile range | When established protocols exist for your organism or assay | Published values may not transfer across laboratories or conditions |
| Resource equation approach | Error degrees of freedom from ANOVA design | Animal studies where power analysis assumptions cannot be met | Does not directly estimate standard deviation, only acceptable sample size ranges |
| Expert opinion or prior knowledge | Historical records or domain experience | When no pilot or published data exist | Subjective and difficult to defend in peer review |

## Understanding Why Standard Deviation Matters in Sample Size Calculations

Sample size calculations in biological research depend on three interconnected quantities: the effect size you wish to detect, the significance level you are willing to accept, and the variability present in your measurements. The standard deviation represents that variability. When you underestimate the standard deviation, your calculated sample size will be too small, and your study will lack the power to detect a real biological effect. When you overestimate it, you waste resources on more animals, more culture plates, or more assay runs than necessary.

The core problem for many researchers is that the standard deviation is rarely known before the experiment begins. You are planning a study precisely because you do not yet have data. This creates a circular challenge: you need the standard deviation to calculate the sample size, but you need data to estimate the standard deviation. The practical solutions involve gathering preliminary information from pilot studies, mining published literature, or using structured estimation approaches that do not require a pre-specified standard deviation.

For researchers working with animal models, the stakes are particularly high. The [Sample Size Calculation in Animal Studies Using Resource Equation Approach](https://pubmed.ncbi.nlm.nih.gov/29386977) paper notes that animal research plays an important role in the pre-clinical phase of clinical trials and that power analysis is the recommended approach. However, when it is not possible to assume the standard deviation and the effect size, the resource equation approach provides an alternative that sets an acceptable range of error degrees of freedom in an analysis of variance. This method acknowledges that sometimes you simply cannot produce a defensible standard deviation estimate before the study begins.

The consequences of getting this wrong extend beyond statistical power. An underpowered study that fails to detect a real effect wastes animal lives and research funding. An overpowered study uses more animals than necessary, raising ethical concerns. Funding agencies and institutional review boards increasingly expect researchers to justify their sample sizes with explicit reasoning about variability. The [National Institutes of Health Grants and Funding](https://grants.nih.gov/) pages describe the expectations for rigorous experimental design in grant applications, and reviewers routinely question sample size justifications that lack a clear basis for the standard deviation estimate.

## Core Principles of Variability Estimation

### Biological Variability Versus Technical Variability

Every measurement you take in a biological experiment contains two distinct sources of variation. Biological variability reflects genuine differences among individual organisms, cells, or samples. Technical variability reflects inconsistencies in your measurement process, including pipetting error, instrument drift, reagent lot differences, and operator technique. The standard deviation you observe in your data combines both sources.

When you estimate the standard deviation for a sample size calculation, you need to think carefully about which type of variability your estimate captures. A published paper that reports a standard deviation from a highly controlled laboratory setting may not reflect the additional technical variability present in your own laboratory. Conversely, a pilot study conducted in your own facility captures both biological and technical variability relevant to your specific conditions, which is why pilot data are often more useful than published values.

The distinction matters for experimental design decisions. If your technical variability is high relative to biological variability, you might reduce the standard deviation by standardizing your protocols, calibrating instruments, or using a single operator for all measurements. If biological variability dominates, you may need to consider whether your experimental units are sufficiently homogeneous or whether you need to control additional factors such as age, sex, genetic background, or environmental conditions.

### The Relationship Between Standard Deviation and Effect Size

Sample size calculations require both an estimate of variability and an estimate of the effect size you wish to detect. These two quantities combine into a standardized effect size, often expressed as the difference between group means divided by the standard deviation. This standardization is what allows researchers to compare studies across different measurement scales.

When you have a small standard deviation, you can detect smaller biological effects with fewer samples. When the standard deviation is large, you need either a larger sample size or a focus on detecting only large effects. This tradeoff is central to experimental planning. A researcher studying a highly variable trait such as behavioral response in outbred animals will need more samples than a researcher measuring a tightly controlled biochemical assay in cultured cells.

The resource equation approach described in the [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977) sidesteps the need to specify an effect size entirely. Instead of asking how many animals are needed to detect a particular effect given the variability, it asks what range of error degrees of freedom produces a statistically valid analysis of variance. This approach is particularly useful when you have no prior data to estimate either the standard deviation or the effect size, which is common in exploratory studies or when working with novel experimental systems.

### Why Small Pilot Samples Produce Unstable Estimates

A standard deviation estimated from a small pilot sample carries substantial uncertainty. The sampling distribution of the variance is wide when the sample size is small, meaning that two pilot studies of the same population could produce very different standard deviation estimates. This instability is a mathematical property of the variance estimator, not a reflection of poor technique.

For example, a pilot study with five observations per group produces a standard deviation estimate that could easily be half or double the true population value. A pilot study with twenty observations per group is more stable but still carries meaningful uncertainty. This uncertainty propagates directly into your sample size calculation. If your standard deviation estimate is 50 percent too low, your calculated sample size will be substantially too small, and your study will be underpowered.

Researchers should therefore treat pilot-derived standard deviations as provisional values and consider using conservative adjustments. Some statisticians recommend using the upper confidence bound of the standard deviation estimate instead of the point estimate itself. This approach builds a margin of safety into the sample size calculation, reducing the risk of an underpowered study at the cost of using more resources than the point estimate would suggest.

## Practical Methods for Estimating Standard Deviation

### Method One: Conduct a Pilot Study

A pilot study is the most direct way to estimate the standard deviation for your specific experimental system. The pilot should use the same protocols, reagents, equipment, and personnel that you will use in the full study. It should also use the same experimental units, whether those are animals, cell culture wells, or field plots.

The size of the pilot study involves a tradeoff. A larger pilot gives you a more stable standard deviation estimate but consumes resources that could go toward the main experiment. A common practical approach is to use approximately 10 to 20 observations per group for the pilot, recognizing that this provides only a rough estimate. The pilot data also serve other purposes, including testing your protocols, identifying technical problems, and estimating the feasibility of your procedures.

When you analyze pilot data, calculate the standard deviation separately for each group instead of pooling all observations. If the standard deviations differ substantially between groups, this itself is important information. Heterogeneous variances may require different sample sizes per group, a transformation of the data, or a statistical analysis that does not assume equal variances.

The pilot study also provides an opportunity to examine the distribution of your data. If the data are heavily skewed, the standard deviation may not be the most appropriate measure of variability. In such cases, you might consider whether a transformation would normalize the data or whether you should plan your analysis around nonparametric methods.

### Method Two: Extract Estimates from Published Literature

When a pilot study is not feasible, published literature can provide standard deviation estimates. The challenge is finding values that are relevant to your specific experimental conditions. A standard deviation reported in a paper using a different animal strain, different housing conditions, different assay protocol, or different laboratory will not perfectly match your situation.

Systematic reviews and meta-analyses are particularly useful sources because they aggregate data across multiple studies. The paper on [estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range](https://pubmed.ncbi.nlm.nih.gov/25524443) addresses a common problem in this context. Many clinical trials and animal studies report their results using the median, minimum and maximum values, or first and third quartiles instead of the mean and standard deviation. To combine results across studies, researchers often need to estimate the sample mean and standard deviation from these summary statistics.

The methods described in that paper are directly applicable to your sample size planning. If you find a published study that reports the median and range for a relevant outcome, you can use these methods to estimate the standard deviation. The paper demonstrates that the proposed estimation methods provide a nearly unbiased estimate of the true sample standard deviation for normal data and a slightly biased estimate for skewed data. This means the methods are useful in practice, but you should be aware that estimates from skewed data carry more uncertainty.

When extracting standard deviation estimates from the literature, document the source carefully. Record the citation, the specific outcome measure, the experimental conditions, and the sample size of the original study. This documentation is essential for defending your sample size calculation in grant applications, institutional review board submissions, and peer-reviewed publications.

### Method Three: Use the Resource Equation Approach

The resource equation approach, described in the [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977), offers a way forward when you cannot estimate the standard deviation or effect size at all. This situation arises frequently in animal research, particularly in exploratory studies, toxicity testing, or when working with novel disease models where no prior data exist.

The approach works through the error degrees of freedom in an analysis of variance. The paper reformulates the error degrees of freedom formulas to guide researchers in calculating the minimum and maximum numbers of animals required. The acceptable range of error degrees of freedom is typically set between 10 and 20, based on statistical reasoning about the stability of variance estimates and the power of the F-test.

To use this approach, you calculate the total number of experimental units minus the number of groups, which gives you the error degrees of freedom. You then adjust the total sample size so that the error degrees of freedom fall within the acceptable range. This method does not require you to specify a standard deviation or effect size, which makes it useful when prior data are completely absent.

The limitation of this approach is that it does not directly estimate the standard deviation. It provides a defensible sample size range based on statistical principles, but it does not tell you the variability you should expect in your data. You should still collect and report the standard deviation from your completed study, both for your own future planning and for the benefit of other researchers who may build on your work.

### Method Four: Combine Expert Opinion with Historical Records

In some situations, you may have access to historical records from your own laboratory or institution. Previous experiments using similar protocols may have recorded standard deviations, even if those experiments addressed different scientific questions. These historical records can be valuable sources of variability estimates because they reflect your specific laboratory conditions.

When historical records are not available, expert opinion can fill the gap. An experienced researcher who has worked with your organism or assay system may have a sense of the typical variability. This approach is subjective, but it can be made more defensible by documenting the reasoning. For example, you might record that the estimate is based on the expert's experience with similar assays and note the range of values they have observed.

The key to using expert opinion effectively is transparency. Reviewers and funding agencies are more likely to accept an expert-based estimate if you clearly state that it is an estimate, explain its basis, and describe how you will verify it. You might also plan an interim analysis to check whether the observed variability matches your estimate, with a pre-specified plan to adjust the sample size if the variability is substantially different.

## Choosing the Right Estimation Method for Your Study

| Study Context | Recommended Method | Rationale |
| --- | --- | --- |
| Established assay in your laboratory | Pilot study with 10 to 20 observations per group | Captures your specific technical and biological variability |
| Novel organism or model system | Resource equation approach | No prior data exist to estimate variability or effect size |
| Well-characterized field with published data | Literature extraction with estimation methods | Published values provide a reasonable starting point |
| Grant application requiring rigorous justification | Pilot study combined with literature comparison | Multiple lines of evidence strengthen the justification |
| Exploratory or hypothesis-generating study | Resource equation approach or expert opinion | Formal power analysis may be premature |

## Step-by-Step Workflow for Estimating Standard Deviation

### Step One: Define Your Primary Outcome and Experimental Design

Before you can estimate the standard deviation, you must know which outcome measure will drive your sample size calculation. Most studies have multiple outcomes, but the sample size is typically calculated for the primary outcome. Choose this outcome before you begin the estimation process and document your choice.

Your experimental design also affects the standard deviation you need. A completely randomized design with independent groups requires a different variability estimate than a paired design or a repeated measures design. In a paired design, the relevant variability is the standard deviation of the differences within pairs, which is often smaller than the standard deviation of the individual measurements. Make sure you are estimating the correct quantity for your design.

### Step Two: Search for Existing Data

Begin by searching the published literature for studies that used similar experimental systems and measured similar outcomes. Systematic reviews and meta-analyses are particularly valuable because they aggregate data across multiple studies. The [NCBI Literature Resources](https://www.ncbi.nlm.nih.gov/) provide access to PubMed and other databases where you can search for relevant studies.

When you find candidate studies, examine their reported summary statistics carefully. If they report the mean and standard deviation, you can use those values directly. If they report the median, range, or interquartile range, you will need to use estimation methods such as those described in the [BMC Medical Research Methodology paper](https://pubmed.ncbi.nlm.nih.gov/25524443) to convert these statistics into an estimated standard deviation.

Document every source you use, including the citation, the outcome measure, the experimental conditions, and the sample size. This documentation will be essential when you justify your sample size calculation to reviewers.

### Step Three: Conduct a Pilot Study if Feasible

If resources permit, conduct a pilot study using your exact protocols and experimental conditions. The pilot should include the same number of groups as your main study and use the same experimental units. Aim for approximately 10 to 20 observations per group, recognizing that this provides a rough estimate.

During the pilot, record any technical problems you encounter. These problems may inflate your variability estimate, and you should address them before the main study. For example, if you notice that your pipetting technique is inconsistent or that your instrument drifts during the day, you should correct these issues and consider whether the pilot data collected before the correction are representative.

Calculate the standard deviation separately for each group in your pilot data. Examine the distributions visually using histograms or box plots. If the data are heavily skewed or contain outliers, consider whether you need to transform the data or use a different analytical approach.

### Step Four: Compare and Synthesize Your Estimates

If you have both pilot data and published literature values, compare them. Substantial discrepancies between your pilot estimate and published values warrant investigation. The discrepancy might reflect differences in experimental conditions, technical variability, or genuine biological differences between your study population and the published population.

When you synthesize multiple estimates, consider using the more conservative value, meaning the larger standard deviation. This approach reduces the risk of an underpowered study. You should also consider the uncertainty in your estimate. If your pilot was small, the standard deviation estimate has wide confidence limits, and you might reasonably use the upper confidence limit in your sample size calculation.

### Step Five: Calculate Your Sample Size

With your standard deviation estimate in hand, you can proceed with the sample size calculation. The specific formula depends on your experimental design and the statistical test you plan to use. For a simple two-group comparison, the calculation requires the standard deviation, the desired effect size, the significance level, and the desired power.

If you are using the resource equation approach because you cannot estimate the standard deviation, calculate the sample size that places your error degrees of freedom within the acceptable range. The [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977) provides the formulas for this calculation.

### Step Six: Document Your Justification

Write a clear justification for your sample size that includes the standard deviation estimate, its source, and the reasoning behind your choice. This justification should be part of your study protocol, grant application, and institutional review board submission. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines that can help you structure your methods section and ensure that you report your sample size justification transparently.

Your documentation should also describe how you will handle the situation if the observed variability in your main study differs substantially from your estimate. A pre-specified plan for this contingency demonstrates rigorous experimental design and can protect you from criticism during peer review.

## Records and Measurements

### What to Record for Standard Deviation Estimation

Maintain a dedicated record for your variability estimation process. This record should include the date of each estimation activity, the method used, the data sources consulted, and the resulting standard deviation estimate. For pilot studies, record the raw data, the protocols used, and any deviations from those protocols.

For literature-based estimates, record the full citation, the outcome measure, the reported summary statistics, the estimation method used to convert those statistics into a standard deviation, and the resulting estimate. The [BMC Medical Research Methodology paper](https://pubmed.ncbi.nlm.nih.gov/25524443) describes several estimation methods, and you should document which one you used and why.

For expert opinion estimates, record who provided the estimate, their relevant experience, the reasoning behind their estimate, and any caveats they expressed. This documentation makes the subjective nature of the estimate transparent and provides a basis for defending it.

### How to Track Variability During the Main Study

Your standard deviation estimate is a prediction about the variability you will observe in your main study. Track the observed variability as data accumulate. Calculate the standard deviation periodically and compare it to your estimate. If the observed standard deviation is substantially larger than your estimate, your study may be underpowered, and you may need to consider increasing your sample size.

Some study designs allow for sample size re-estimation at an interim point. This approach uses the observed variability from the first portion of the study to recalculate the required sample size for the remainder. If you plan to use this approach, specify the interim analysis plan in advance, including the criteria for adjusting the sample size and the statistical methods for the final analysis.

### Using Completed Study Data for Future Planning

When your study is complete, record the observed standard deviation for each outcome measure. This information is valuable for your own future studies and for other researchers in your field. The [NIH Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy) describes expectations for sharing research data, and including variability information in your shared data enhances its utility.

Consider publishing your variability estimates, particularly if you are working in an area where such data are scarce. A paper that reports standard deviations for common outcome measures in a particular animal model or assay system can be highly cited and valuable to the research community.

## Common Failure Patterns in Standard Deviation Estimation

### Underestimating Variability from Small Pilot Samples

The most common failure pattern is using a standard deviation estimate from a very small pilot sample without accounting for the uncertainty in that estimate. A pilot with five observations per group produces a standard deviation estimate that is highly unstable. If that estimate happens to be low, the calculated sample size will be too small, and the main study will be underpowered.

The solution is to recognize the instability of small-sample variance estimates and to use conservative adjustments. Consider using the upper confidence bound of the standard deviation instead of the point estimate. Alternatively, plan an interim analysis to check the observed variability and adjust the sample size if needed.

### Assuming Published Values Transfer Across Conditions

Published standard deviations are often treated as universal constants, but they are specific to the conditions under which they were measured. A standard deviation reported for a particular mouse strain in one laboratory may not apply to the same strain in a different facility with different housing, diet, microbiome, or personnel.

When using published values, consider whether the conditions match your own. If they do not match, the published value may be an underestimate or overestimate of your variability. When in doubt, use a larger standard deviation to build in a margin of safety.

### Confusing the Standard Deviation of the Population with the Standard Error of the Mean

Standard error (SE) quantifies the sampling precision of an estimate: SE = σ/√n, where σ is the population standard deviation and n is the sample size.

Check your formulas carefully. Sample size calculations require the standard deviation of the outcome measure, not the standard error. If you are extracting values from published papers, verify whether the reported value is a standard deviation or a standard error. The notation in the paper should make this clear, but errors in reporting do occur.

### Ignoring the Distribution of the Data

The standard deviation is most meaningful when the data are approximately normally distributed. If your outcome measure is heavily skewed, the standard deviation may not adequately capture the variability, and your sample size calculation may be misleading.

Examine the distribution of your pilot data or published data before proceeding. If the data are skewed, consider whether a transformation would normalize them. If you plan to use nonparametric methods, the sample size calculation requires different considerations, and you may need to consult a statistician.

### Failing to Document the Basis for the Estimate

A sample size justification that does not explain the basis for the standard deviation estimate is difficult to defend. Reviewers may reject the justification as arbitrary, and the study may be criticized for using an unjustified sample size.

Document every aspect of your estimation process. Record the sources you consulted, the methods you used, the assumptions you made, and the reasoning behind your choices. This documentation demonstrates rigorous experimental design and protects you during peer review.

## Limitations and Interpretation

### The Uncertainty in Any Standard Deviation Estimate

Every standard deviation estimate carries uncertainty, regardless of the method used to produce it. Pilot studies produce estimates with wide confidence intervals when the pilot sample is small. Literature-based estimates inherit the uncertainty of the original studies and add the uncertainty of the estimation methods used to convert summary statistics. Expert opinion estimates carry the uncertainty of subjective judgment.

This uncertainty should be acknowledged in your sample size justification. instead of presenting a single standard deviation value, consider presenting a range of plausible values and showing how the required sample size changes across that range. This sensitivity analysis demonstrates that you understand the uncertainty in your estimate and have planned accordingly.

### The Resource Equation Approach Does Not Estimate Standard Deviation

The resource equation approach provides a defensible sample size range without requiring a standard deviation estimate, but it does not tell you what variability to expect. You will still need to measure and report the standard deviation from your completed study. This information is valuable for future studies and for the broader research community.

The [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977) recommends the power analysis approach when possible and presents the resource equation approach as an alternative when the standard deviation and effect size cannot be assumed. This framing is important. The resource equation approach is not a substitute for power analysis when you have the information to conduct one. It is a fallback for situations where the information is genuinely unavailable.

### Estimation Methods for Skewed Data

The methods for estimating the standard deviation from median, range, and interquartile range perform well for normally distributed data but are less accurate for skewed data. The [BMC Medical Research Methodology paper](https://pubmed.ncbi.nlm.nih.gov/25524443) notes that the proposed methods provide a nearly unbiased estimate for normal data and a slightly biased estimate for skewed data.

If you suspect your outcome measure is skewed, be cautious in using these estimation methods. Consider whether a transformation would normalize the data or whether you should use a different approach to sample size planning. Consulting a statistician is advisable when you are working with skewed data and have limited prior information.

### The Role of Reporting Guidelines

The EQUATOR Network maintains a comprehensive, searchable registry of reporting guidelines spanning clinical trials, observational studies, systematic reviews, and qualitative research.

Following reporting guidelines also contributes to the broader research ecosystem. When you report your standard deviation estimates and sample size calculations transparently, other researchers can build on your work and plan their own studies more effectively.

## Welfare and Safety Context

### Ethical Considerations in Animal Studies

The estimation of variability and the resulting sample size calculation have direct ethical implications in animal research. An underpowered study uses animals without a reasonable chance of detecting a real effect, wasting animal lives and potentially leading to inconclusive results that cannot inform future research. An overpowered study uses more animals than necessary, causing avoidable harm.

The [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977) emphasizes the importance of calculating the minimum and maximum numbers of animals required in animal research. The resource equation approach provides a framework for this calculation when power analysis is not feasible. Researchers should also consult institutional animal care and use committees, which review sample size justifications as part of their ethical oversight.

### Biosafety Considerations in Laboratory Work

Pilot studies and main experiments involving biological samples require appropriate biosafety practices. The [WHO Laboratory Biosafety Manual](https://www.who.int/publications/i/item/9789240011311) provides guidance on safe handling of biological materials. When you are conducting pilot studies to estimate variability, follow the same biosafety protocols you will use in the main study.

Biosafety considerations can also affect your variability estimates. If you are working with hazardous materials that require special handling, the additional complexity of the procedures may increase technical variability. Your pilot study should reflect the actual conditions of the main study, including any biosafety constraints.

### Data Management and Sharing

The [NIH Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy) describes expectations for managing and sharing research data. These expectations apply to the data you collect during pilot studies and main experiments. Plan your data management approach before you begin, including how you will store, organize, and share your data.

Sharing your variability data contributes to the research community by helping other researchers plan their studies. When you publish your results, include the standard deviations for your outcome measures and describe the conditions under which they were measured. This information is as valuable as the primary findings of your study.

### Research Integrity and Publication Ethics

The [Committee on Publication Ethics Core Practices](https://publicationethics.org/core-practices) describe expectations for authorship, peer review, data integrity, and conflict of interest. These practices apply to your sample size planning and reporting. You should report your sample size justification honestly, including the limitations of your standard deviation estimate.

If you adjust your sample size during the study based on interim variability estimates, report this adjustment transparently. The [EQUATOR Network](https://www.equator-network.org/) reporting guidelines can help you structure this reporting. Transparent reporting of sample size decisions, including any deviations from the original plan, is essential for the integrity of your research.

## Professional Escalation Criteria

### When to Consult a Statistician

You should consult a statistician when your experimental design is complex, your data are likely to be skewed, or you are uncertain about the appropriate sample size calculation. Specific situations that warrant statistical consultation include:

- Studies with multiple primary outcomes or complex designs such as factorial, split-plot, or hierarchical designs
- Outcome measures with known non-normal distributions or heavy skewness
- Studies requiring interim analysis or adaptive sample size re-estimation
- Situations where the standard deviation estimate is highly uncertain and the consequences of an underpowered study are severe

A statistician can help you choose the appropriate estimation method, conduct sensitivity analyses, and develop a defensible sample size justification.

### When to Consult an Institutional Animal Care and Use Committee

Institutional animal care and use committees review sample size justifications as part of their ethical oversight of animal research. You should consult your committee early in the planning process, particularly if you are using the resource equation approach or if your standard deviation estimate is based on limited data.

The committee can provide guidance on acceptable sample size ranges and may require additional justification for studies that use unusually large or small numbers of animals. Engaging the committee early can prevent delays in study approval.

### When to Seek Additional Data

If your standard deviation estimate is based on very limited data and the consequences of an underpowered study are severe, consider whether additional pilot data are feasible. A slightly larger pilot study can substantially improve the stability of your standard deviation estimate.

You should also consider whether published data from other laboratories can supplement your own pilot data. Combining multiple sources of information can provide a more robust estimate than any single source.

## Frequently Asked Questions

### What is the simplest way to estimate the standard deviation when I have no prior data?

The resource equation approach is the most practical method when you have no prior data to estimate the standard deviation or effect size. This approach, described in the [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977), uses the error degrees of freedom in an analysis of variance to determine an acceptable range of sample sizes. It does not require you to specify a standard deviation, which makes it useful for exploratory studies or novel experimental systems.

### How many observations should I include in a pilot study to estimate the standard deviation?

A pilot study with approximately 10 to 20 observations per group provides a rough estimate of the standard deviation. Smaller pilots produce highly unstable estimates, while larger pilots consume resources that could go toward the main study. Recognize that even a 20-observation pilot provides only a provisional estimate, and consider using the upper confidence bound of the standard deviation in your sample size calculation.

### Can I use the standard deviation from a published paper that used a different animal strain or assay protocol?

You can use published values as a starting point, but you should consider whether the conditions match your own. Differences in animal strain, housing, diet, assay protocol, or laboratory can substantially affect variability. When in doubt, use a larger standard deviation to build in a margin of safety, and document the source of your estimate in your sample size justification.

### How do I estimate the standard deviation when a paper only reports the median and range?

The paper on [estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range](https://pubmed.ncbi.nlm.nih.gov/25524443) describes methods for converting these summary statistics into an estimated standard deviation. These methods perform well for normally distributed data and are slightly biased for skewed data. Document which estimation method you used and be cautious when applying the methods to skewed outcomes.

### What is the difference between the standard deviation and the standard error in sample size calculations?

The standard error of the mean decreases with increasing sample size, scaling as the standard deviation divided by the square root of n - the statistical basis for the law of large numbers.

### When should I use the resource equation approach instead of power analysis?

Use the resource equation approach when you cannot assume the standard deviation and the effect size, as described in the [animal studies paper](https://pubmed.ncbi.nlm.nih.gov/29386977). This situation arises in exploratory studies, novel model systems, or when no prior data exist. When you have reasonable estimates of the standard deviation and effect size, power analysis is the recommended approach.

### How should I document my standard deviation estimate for a grant application?

Document the source of your estimate, the method used to produce it, and the reasoning behind your choices. For pilot data, include the raw data and protocols. For literature-based estimates, include the full citation and the estimation method used. For expert opinion, record who provided the estimate and their relevant experience. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines that can help you structure your methods section.

### What should I do if the observed variability in my main study is much larger than my estimate?

If the observed variability is substantially larger than your estimate, your study may be underpowered. Consider whether your study design allows for sample size re-estimation at an interim point. If you planned for this contingency in advance, you can adjust your sample size using the observed variability. If not, you may need to consult a statistician to determine the best course of action.

## Related Diagnostic Guides

- [RNA Sequencing Protocol: From Sample to Data](/knowledge/diagnostics/molecular/rna-sequencing-protocol-from-sample-to-data)
- [Standard Plate Count: Formula, Calculation, and Reporting](/knowledge/diagnostics/microbiology/standard-plate-count-formula-calculation-and-reporting)
- [Animal Biology](/blog/guides/animal-biology)

## References and Further Reading

- [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books). National Library of Medicine.
- [EQUATOR Network](https://www.equator-network.org/). EQUATOR Network.
- [Core Practices](https://publicationethics.org/core-practices). Committee on Publication Ethics.
- [NIH Grants and Funding](https://grants.nih.gov/). National Institutes of Health.
- [ORCID for Researchers](https://info.orcid.org/researchers). ORCID.
- [Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy). National Institutes of Health.
- [Laboratory Biosafety Manual](https://www.who.int/publications/i/item/9789240011311). World Health Organization.
- [NCBI Literature Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information.
- [Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range.](https://pubmed.ncbi.nlm.nih.gov/25524443). BMC medical research methodology, 2014.
- [Sample Size Calculation in Animal Studies Using Resource Equation Approach.](https://pubmed.ncbi.nlm.nih.gov/29386977). The Malaysian journal of medical sciences : MJMS, 2017.
- [Sample Size for Assessing Agreement between Two Methods of Measurement by Bland-Altman Method.](https://pubmed.ncbi.nlm.nih.gov/27838682). The international journal of biostatistics, 2016.

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