# Designing a Factorial Experiment with Multiple Factors

## Quick Answer

- Use a factorial design template to specify every factor, level, and combination before data collection, which strengthens grant proposals and protocol clarity.
- Structure the template around five core sections: factors and levels, randomization, replication, blocking, and the statistical analysis plan.
- A factorial template cannot replace consultation with a biostatistician, especially for sample size calculations and interpretation of interaction effects.

## At a Glance

| Template Section | Purpose | Common Omission | Consequence of Omission |
|---|---|---|---|
| Factors and Levels | Define each independent variable and its discrete values | Unclear level definitions or continuous variables treated as categorical | Ambiguous treatment groups and irreproducible protocols |
| Randomization | Assign experimental units to treatment combinations without bias | No randomization scheme or unit allocation description | Systematic error and invalid statistical inference |
| Replication | Specify the number of independent experimental units per combination | Confusing technical replicates with biological replicates | Overstated precision and false confidence in results |
| Analysis Plan | State the statistical model, comparisons, and software | No prespecified model or interaction terms | Data dredging and inflated false positive rates |

## Defining the Factorial Design in a Grant Proposal

A factorial experiment tests two or more factors simultaneously, with every level of one factor combined with every level of the other factors. This structure allows researchers to examine both the individual effect of each factor and the interaction between factors. For a life science grant proposal, the factorial design must be described with enough specificity that a reviewer can assess the scientific rationale, the feasibility of the sample size, and the validity of the planned statistical analysis.

The National Library of Medicine provides access to authoritative biomedical books and research method references that describe the principles of experimental design, including randomization, replication, and control. These resources serve as a gateway for researchers who need to verify the methodological basis of their design choices before submitting a proposal.

The factorial design is distinct from a one-factor-at-a-time approach. In a one-factor-at-a-time experiment, the researcher varies one factor while holding all others constant. This approach cannot detect interactions between factors and requires more experimental runs to achieve the same level of precision. A factorial design is more efficient because each observation contributes information about multiple factors and their interactions.

## Core Principles of Factorial Design

### Factors and Levels

A factor is an independent variable that the researcher manipulates. Each factor has two or more discrete levels. For example, a cell culture experiment might have the factor temperature with levels of 37 degrees Celsius and 39 degrees Celsius, and the factor glucose concentration with levels of 5 millimolar and 25 millimolar. The complete factorial design includes all four combinations of these two factors.

The choice of levels is a scientific decision that must be justified in the proposal. Levels should be selected based on prior literature, pilot data, or a clear biological rationale. The researcher must state whether the levels represent a range of interest or a comparison of specific conditions. If the levels are chosen arbitrarily, the results may not generalize to the broader range of conditions that the researcher intends to study.

### The Complete Factorial Design

In a complete factorial design, the researcher measures every possible combination of factor levels. For two factors with two levels each, this yields four treatment combinations. For three factors with two levels each, the design has eight treatment combinations. The number of treatment combinations grows multiplicatively, which is the primary limitation of the factorial approach.

The complete factorial design is the most straightforward to describe in a grant proposal because the treatment structure is explicit. The researcher can list every treatment combination in a table, which helps reviewers understand the scope of the experiment. The design also allows the researcher to estimate all main effects and all interactions without confounding.

### Main Effects and Interactions

A main effect is the average effect of one factor across all levels of the other factors. An interaction occurs when the effect of one factor depends on the level of another factor. For example, a drug might reduce cell viability at one temperature but have no effect at another temperature. This is an interaction between the drug factor and the temperature factor.

The ability to detect interactions is the primary reason to use a factorial design. If the researcher expects that the effect of one factor will depend on another factor, a factorial design is necessary. The statistical analysis plan must include terms for the interactions, and the sample size must be adequate to detect these effects.

## The Factorial Design Template

### Section 1: Study Title and Objective

The template begins with a clear statement of the study objective. This statement should describe the biological question in one or two sentences. The objective should be specific enough that a reviewer can determine whether the factorial design is appropriate for the question.

The objective should also state the primary outcome variable. This is the measurement that will be used to evaluate the effects of the factors. The primary outcome must be defined before the experiment begins, and it should be measured consistently across all treatment combinations.

### Section 2: Factors and Levels Table

The factors and levels table is the core of the template. Each row lists one factor, its levels, and the rationale for the level selection. The table should also indicate whether each factor is fixed or random. A fixed factor has levels that are the only ones of interest, while a random factor has levels that are a sample from a larger population.

The table should include the units of measurement for each level. For example, a temperature factor should list the levels in degrees Celsius, and a concentration factor should list the levels in millimolar. This prevents ambiguity in the protocol and ensures that the experiment can be reproduced.

### Section 3: Experimental Unit and Replication

The experimental unit is the smallest unit to which a treatment is independently applied. In a cell culture experiment, the experimental unit is the individual well or flask, not the individual cell. In an animal study, the experimental unit is the individual animal, not the individual measurement.

The template must state the number of experimental units per treatment combination. This is the replication number. The replication number is distinct from the number of technical replicates, which are repeated measurements on the same experimental unit. Technical replicates do not increase the statistical power for detecting treatment effects.

### Section 4: Randomization Scheme

The randomization scheme describes how experimental units are assigned to treatment combinations. The template should state the method of randomization, such as a random number generator or a random permutation. The scheme should also describe any blocking or stratification that is used to control for known sources of variability.

Randomization is essential for valid statistical inference. Without randomization, the treatment groups may differ systematically in ways that are not accounted for in the analysis. The template should include a description of the randomization procedure so that a reviewer can assess whether the design is unbiased.

### Section 5: Statistical Analysis Plan

The analysis plan states the statistical model that will be used to analyze the data. For a factorial design, the model typically includes the main effects of each factor and the interaction terms. The plan should also state the significance level, the software that will be used, and the method for handling missing data.

The analysis plan should be written before the data are collected. This prevents the researcher from choosing an analysis after seeing the results, which can inflate the false positive rate. The plan should also state the primary comparison of interest, which is the comparison that the sample size is designed to detect.

## Practical Implementation Steps

### Step 1: Define the Biological Question

The first step is to write the biological question in a single sentence. The question should identify the factors, the outcome, and the population of interest. For example, the question might be, "Does the effect of drug concentration on cell viability depend on the temperature of the culture?"

This question is the basis for the entire design. If the question is vague, the design will be vague. The researcher should revise the question until it is specific enough that a reviewer can determine the appropriate design.

### Step 2: Select the Factors and Levels

The researcher should list all factors that are expected to affect the outcome. The number of factors should be limited to those that are essential to the question. Each factor should have a small number of levels, typically two or three. The levels should be chosen to represent the range of conditions that are relevant to the biological question.

The researcher should also consider whether the factors are expected to interact. If the researcher has no reason to expect an interaction, a factorial design may still be useful because it is efficient, but the interaction terms may not be the primary focus.

### Step 3: Determine the Experimental Unit

The experimental unit must be identified before the sample size is calculated. The researcher should ask, "What is the smallest unit that receives the treatment independently?" The answer to this question determines the replication number.

In a cell culture experiment, the experimental unit is the well or the plate, depending on how the treatment is applied. If the treatment is applied to the entire plate, the plate is the experimental unit. If the treatment is applied to each well, the well is the experimental unit.

### Step 4: Calculate the Sample Size

The sample size is the number of experimental units per treatment combination. The sample size should be calculated based on the expected effect size, the variability, and the desired power. The researcher should consult a biostatistician for this calculation, because the calculation depends on the specific design and the statistical model.

The sample size calculation should be described in the grant proposal. The description should include the effect size that the study is designed to detect, the significance level, and the power. The researcher should also state the software that was used for the calculation.

### Step 5: Write the Randomization Scheme

The randomization scheme should be written before the experiment begins. The researcher should describe how the experimental units will be assigned to the treatment combinations. The scheme should include the use of a random number generator and any blocking that will be used.

The randomization scheme should be recorded in the laboratory notebook. This record is important for the reproducibility of the experiment and for the audit of the data.

### Step 6: Write the Analysis Plan

The analysis plan should state the statistical model, the primary comparison, and the method for handling missing data. The plan should be written before the data are collected. The researcher should also state the software that will be used for the analysis.

The analysis plan should be included in the grant proposal. The reviewer will assess whether the plan is appropriate for the design and whether the sample size is adequate.

## Records and Measurements

### Laboratory Notebook Documentation

The laboratory notebook should contain the randomization scheme, the treatment assignments, and the dates of the experiment. The notebook should also contain the raw data, the calculations, and the statistical analysis. The notebook is the primary record of the experiment and should be complete enough for another researcher to reproduce the work.

The notebook should also record any deviations from the protocol. If a well is contaminated or an animal is lost, the deviation should be noted. The deviation should be described in the results and considered in the analysis.

### Data Collection and Storage

The data should be collected in a structured format that matches the treatment plan. Each row of the data table should represent one experimental unit, and each column should represent one variable. The data should be stored in a secure location and backed up regularly.

The data management plan should be described in the proposal. The plan should state how the data will be stored, who will have access, and how the data will be shared. The National Institutes of Health provides guidance on data management and sharing expectations for funded research.

### Metadata and Annotations

The metadata should describe the experimental conditions, the equipment, and the software that were used. The metadata should be sufficient for another researcher to understand the context of the data. The annotations should include the date, the batch of reagents, and the calibration of the instruments.

The metadata should be stored with the data. The researcher should use a consistent naming convention for the files and the variables. This makes the data easier to analyze and to share.

## Common Failure Patterns

### Confusing Technical and Biological Replicates

A common failure is to treat repeated measurements on the same experimental unit as independent replicates. This overstates the sample size and the statistical power. The researcher should identify the experimental unit and count only the independent units in the sample size.

For example, if a researcher measures the same well multiple times, the measurements are technical replicates. The researcher should average the technical replicates and treat the well as the experimental unit. The sample size is the number of wells, not the number of measurements.

### Ignoring Interactions in the Analysis

A second failure is to analyze the data with a model that includes only the main effects. This ignores the interaction terms and can lead to incorrect conclusions. The analysis plan should include the interaction terms, and the researcher should test whether the interactions are significant.

If the interaction is significant, the main effects should be interpreted with caution. The effect of one factor depends on the level of the other factor, so a single main effect does not describe the relationship.

### Unbalanced Designs

A third failure is an unbalanced design, where the number of replicates is not the same for all treatment combinations. An unbalanced design is more difficult to analyze and can reduce the power of the test. The researcher should plan for equal replication, and if the design becomes unbalanced due to loss, the analysis should account for the imbalance.

### Incomplete Randomization

A fourth failure is to assign treatments without randomization. This can introduce systematic bias into the experiment. The researcher should use a random number generator to assign the treatments and should record the assignment in the notebook.

## Limitations of the Factorial Template

### The Template Does Not Replace Statistical Consultation

The template is a tool for organizing the design, but it does not replace the advice of a biostatistician. The sample size calculation, the choice of the statistical model, and the interpretation of the interactions require statistical expertise. The researcher should consult a biostatistician before the proposal is submitted.

### The Template Does Not Address All Design Types

The template is designed for a complete factorial design with fixed factors. It does not address fractional factorial designs, nested designs, or split-plot designs. These designs are more complex and require a different template.

### The Template Does Not Guarantee Reproducibility

The template helps the researcher to describe the design, but it does not guarantee that the experiment is reproducible. Reproducibility requires careful documentation of the protocol, the reagents, and the equipment. The researcher should follow the reporting guidelines that are relevant to the study type.

## Reporting Guidelines and Transparency

The EQUATOR Network provides a collection of reporting guidelines for health research. These guidelines help researchers to report their methods and results transparently. The researcher should select the guideline that is relevant to the study type and follow the checklist.

The reporting guideline should be cited in the proposal and in the final report. The checklist should be completed and submitted with the manuscript. This improves the transparency of the research and helps the reviewer to assess the quality of the study.

The Committee on Publication Ethics (COPE) provides core practices for publication ethics, including authorship, peer review, and data. The researcher should be aware of these practices and should follow them in the preparation of the manuscript. The researcher should also ensure that the data are reported accurately and that the results are not misrepresented.

## Researcher Identity and Data Management

The researcher should maintain an ORCID record that is linked to the research outputs. The ORCID record provides a persistent identifier for the researcher and helps to connect the researcher to their publications and data. The researcher should update the record when the research is published.

The data management plan should be described in the proposal. The plan should state how the data will be stored, shared, and preserved. The NIH Data Management and Sharing Policy describes the expectations for funded research. The researcher should follow the policy and should describe the plan in the proposal.

## Safety and Regulatory Context

The factorial design does not change the biosafety requirements of the experiment. The researcher should follow the institutional biosafety guidelines and the relevant regulations. The researcher should also follow the guidelines for the use of animals or human subjects, if the study involves these.

The researcher should be aware of the ethical requirements for the study. The study should be approved by the institutional review board or the institutional animal care and use committee before the experiment begins. The approval should be documented in the proposal.

## Professional Escalation Criteria

The researcher should consult a biostatistician when the design is complex, when the sample size is difficult to calculate, or when the analysis plan is not clear. The researcher should also consult a biostatistician when the data are not balanced or when the interaction terms are difficult to interpret.

The researcher should consult the institutional research office when the study involves human subjects, animals, or hazardous materials. The research office can provide guidance on the regulatory requirements and the approval process.

The researcher should consult the institutional data management office when the data are sensitive or when the data sharing requirements are not clear. The data management office can provide guidance on the data management plan and the data repository.

## Building a Factorial Design Review Checklist for Grant Submission

A factorial design template helps you write a proposal, but it does not guarantee that a reviewer will understand the design or that the proposal will survive a rigorous review. Grant reviewers often evaluate factorial designs by checking whether the proposal answers a specific set of questions about the experimental structure. You can use the same set of questions to review your own proposal before submission. This section provides a practical decision framework that converts the template sections into a review checklist, a scoring system for design completeness, and a troubleshooting method for common reviewer objections.

### The Review Checklist Framework

The checklist framework translates each template section into a question that a reviewer would likely ask. For each question, you should be able to point to a specific sentence, table, or figure in your proposal that provides the answer. If you cannot point to the answer, the proposal is incomplete.

The framework has seven questions, each mapped to a template section. The first question addresses the scientific rationale. The reviewer will ask whether the factorial design is necessary for the stated objective. You should state explicitly why a factorial design is preferred over a one-factor-at-a-time approach. The answer should mention the interaction hypothesis. If you do not expect an interaction, you should justify why the factorial design is still efficient for your question.

The second question addresses the factor structure. The reviewer will ask whether each factor has a clear definition, a stated unit of measurement, and a justification for the selected levels. The factors and levels table in the template should answer this question. The table must include the units for each level and the rationale for the level selection. If a factor is continuous, such as temperature or concentration, the reviewer will ask why you selected specific discrete levels instead of a continuous range.

The third question addresses the experimental unit. The reviewer will ask whether the experimental unit is clearly identified and whether the replication number is consistent with that identification. This is the most common source of reviewer objections. The proposal should state the experimental unit in a single sentence and then state the number of experimental units per treatment combination. The proposal should also distinguish biological replicates from technical replicates.

The fourth question addresses randomization. The reviewer will ask how the experimental units are assigned to the treatment combinations and whether any blocking or stratification is used. The proposal should describe the randomization procedure and the blocking scheme. The description should be specific enough that another researcher could reproduce the assignment.

The fifth question addresses the statistical analysis. The reviewer will ask whether the analysis plan includes the main effects and interaction terms, whether the primary comparison is stated, and whether the sample size is adequate. The proposal should state the statistical model and the primary comparison. The sample size calculation should be described with the effect size, significance level, and power.

The sixth question addresses feasibility. The reviewer will ask whether the total number of experimental units is realistic for the laboratory, the budget, and the timeline. The proposal should state the total number of experimental units and the number of treatment combinations. The total should be consistent with the resources described in the budget and the personnel plan.

### Scoring the Design Completeness

You can score your proposal against the six questions using a simple three-level scale for each question. A score of zero means the proposal does not answer the question. A score of one means the proposal provides a partial answer. A score of two means the proposal provides a complete answer with a specific description that a reviewer can verify.

A total score of ten or higher out of twelve indicates that the proposal is likely to pass a design review. A score of eight or nine indicates that the proposal has gaps that a reviewer might question. A score of seven or lower indicates that the proposal needs substantial revision before submission.

The scoring system is not a substitute for a biostatistician review. It is a self-assessment tool that helps you identify the weakest sections of the proposal before you send it to a consultant or a collaborator. The scoring system also helps you prioritize your revision time. If the experimental unit question scores zero, you should fix that section before you refine the analysis plan.

### The Decision Tree for Design Choices

The decision framework includes a decision tree that helps you choose the correct design structure for your question. The tree starts with the question of whether you need to test multiple factors simultaneously. If you only need to test one factor, a factorial design is not necessary. If you need to test two or more factors, the tree asks whether you expect an interaction between the factors.

If you expect an interaction, the tree directs you to a complete factorial design. If you do not expect an interaction, the tree asks whether the number of treatment combinations is feasible. If the number is feasible, the complete factorial design is still the best choice because it is efficient and allows you to check for unexpected interactions. If the number is not feasible, the tree directs you to consider a fractional factorial design or a different approach.

The tree also addresses the choice of levels. If a factor is continuous and the relationship with the outcome is expected to be linear, two levels may be sufficient. If the relationship is expected to be nonlinear, the tree directs you to use three or more levels. The tree also asks whether the factor is fixed or random. A fixed factor has levels that are the only ones of interest. A random factor has levels that are a sample from a larger population. The choice affects the statistical model and the interpretation of the results.

### Troubleshooting Common Reviewer Objections

The troubleshooting method addresses the most common reviewer objections to factorial designs. The first objection is that the design has too many treatment combinations. The reviewer may say that the proposal is not feasible because the number of combinations is too large. The response is to show the total number of experimental units and the resources required. If the total is too large, the proposal should describe a reduction strategy, such as removing a factor, reducing the number of levels, or using a fractional factorial design.

The second objection is that the sample size is not justified. The reviewer may say that the proposal does not state the effect size or the power. The response is to provide the sample size calculation with the effect size, the variability estimate, the significance level, and the power. The calculation should be based on the primary comparison, which is the comparison that the study is designed to detect.

The third objection is that the experimental unit is not clear. The reviewer may say that the proposal does not distinguish between the experimental unit and the measurement unit. The response is to state the experimental unit in a single sentence and to state the number of experimental units per treatment combination. The proposal should also state the number of technical replicates and explain that technical replicates are averaged before the analysis.

The fourth objection is that the randomization is not described. The reviewer may say that the proposal does not explain how the treatment assignments are made. The response is to describe the randomization procedure, including the use of a random number generator and any blocking scheme. The response should also state that the randomization scheme is recorded in the laboratory notebook.

The fifth objection is that the analysis plan is not prespecified. The reviewer may say that the proposal does not state the statistical model or the primary comparison. The response is to state the model with the main effects and interaction terms, the primary comparison, and the method for handling missing data. The response should also state the software that will be used.

### The Decision Log for Design Revisions

A decision log is a record of the design choices and the reasons for those choices. The log is useful when you revise the proposal after a reviewer comment or when you discuss the design with a biostatistician. The log has four columns. The first column lists the design element, such as the number of factors, the number of levels, or the experimental unit. The second column lists the choice that was made. The third column lists the reason for the choice. The fourth column lists the date of the decision and the person who made the decision.

The decision log is not a substitute for the laboratory notebook. The laboratory notebook records the execution of the experiment. The decision log records the design decisions and the rationale. The decision log is useful for the final report because it explains why the design has the structure that it does.

The decision log also helps when the proposal is revised after a reviewer comment. If a reviewer asks for a change to the design, the log shows the original decision and the reason. The researcher can then decide whether the reviewer comment changes the reason and whether the design should be changed. The log prevents the researcher from making a change without understanding the consequence.

### The Pre-Submission Review Protocol

The pre-submission review protocol is a structured process for checking the proposal before submission. The protocol has four steps. The first step is to complete the scoring checklist. The second step is to identify the sections that score one or lower. The third step is to revise those sections. The fourth step is to have a colleague or a biostatistician review the revised proposal.

The protocol should be completed at least one week before the submission deadline. This allows time for the revision and for the review by a colleague. The protocol should not be completed on the day of the submission because the researcher will not have time to address the findings.

The protocol also includes a check of the reporting guidelines. The EQUATOR Network provides a collection of reporting guidelines for health research. The researcher should select the guideline that is relevant to the study type and complete the checklist. The checklist should be included in the proposal or referenced in the methods section.

### The Record System for Design Documentation

The record system for design documentation has three components. The first component is the decision log, which records the design choices and the rationale. The second component is the randomization record, which records the assignment of experimental units to treatment combinations. The third component is the analysis plan record, which records the statistical model and the primary comparison.

The randomization record should include the seed for the random number generator, the date of the assignment, and the person who performed the assignment. The seed allows the assignment to be reproduced. The record should also include the blocking scheme and the assignment of blocks to the experimental units.

The analysis plan record should include the statistical model, the primary comparison, the significance level, and the software version. The record should also include the method for handling missing data and the method for checking the assumptions of the model. The record should be dated and signed by the researcher.

The record system should be stored in a secure location and backed up regularly. The record system should be described in the data management plan. The National Institutes of Health provides guidance on data management and sharing expectations for funded research. The data management plan should state how the design records will be stored and preserved.

### The Comparison of the Design Structure

The comparison of the design structure is a method for checking whether the proposal describes the design consistently across all sections. The method has three steps. The first step is to list the factors and levels from the factors and levels table. The second step is to list the treatment combinations that are described in the randomization section. The third step is to compare the two lists.

The comparison should show that the number of treatment combinations in the randomization section matches the number of combinations in the factors and levels table. If the numbers do not match, the proposal has an error. The error should be corrected before submission.

The comparison should also show that the sample size is consistent with the number of treatment combinations and the number of replicates. The total number of experimental units should be the product of the number of treatment combinations and the number of replicates. If the total is not the product, the proposal has an error.

The comparison should also show that the analysis plan includes the interaction terms for all factors. If the analysis plan does not include the interaction terms, the proposal has an error. The error should be corrected before submission.

### The Professional Escalation Criteria for Design Review

The professional escalation criteria identify the situations in which the researcher should consult a biostatistician or a research methodologist. The first criterion is the use of a fractional factorial design. The fractional factorial design is complex and requires statistical expertise to select the appropriate fraction and to interpret the confounding.

The second criterion is the use of a split-plot design. The split-plot design has two levels of experimental units and requires a specific analysis. The researcher should consult a biostatistician if the design is a split-plot.

The third criterion is the use of a nested design. The nested design has factors that are nested within other factors. The nested design requires a specific analysis and the researcher should consult a biostatistician.

The fourth criterion is the loss of experimental units. If the design becomes unbalanced because of the loss of units, the researcher should consult a biostatistician. The analysis of an unbalanced design is more complex than the analysis of a balanced design.

The fifth criterion is the difficulty of the sample size calculation. If the researcher cannot calculate the sample size or does not know the effect size, the researcher should consult a biostatistician. The sample size calculation is critical for the validity of the study.

### The Review of the Design by a Colleague

The review of the design by a colleague is a method for catching errors that the researcher cannot see. The colleague should be someone who is familiar with the experimental system but is not involved in the study. The colleague should review the proposal against the six questions in the checklist.

The colleague should also review the decision log and the randomization record. The colleague should check that the design is consistent across all sections. The colleague should provide written feedback on the proposal.

The review by a colleague should be completed before the proposal is submitted. The review should be documented in the decision log. The researcher should respond to the colleague's feedback and record the response in the decision log.

### The Final Check Before Submission

The final check before submission is a list of the items that should be present in the proposal. The first item is the statement of the objective and the primary outcome. The second item is the factors and levels table with the units and the rationale. The third item is the statement of the experimental unit and the replication number. The fourth item is the description of the randomization scheme. The fifth item is the statistical analysis plan with the model and the primary comparison. The sixth item is the sample size calculation with the effect size, the power, and the significance level.

The final check also includes the reporting guideline checklist and the data management plan. The reporting guideline checklist should be completed and referenced in the methods section. The data management plan should describe the storage, sharing, and preservation of the data.

The final check should be completed by the researcher and by the colleague. The researcher should sign the final check and the colleague should sign the final check. The signed final check should be stored with the decision log and the proposal.

## Frequently Asked Questions

### What is the difference between a factorial design and a one-factor-at-a-time design?

A factorial design tests all combinations of the factors, while a one-factor-at-a-time design varies one factor while holding the others constant. The factorial design can detect interactions and is more efficient.

### How many experimental units do I need for a factorial design?

The number of experimental units depends on the number of factors, the number of levels, and the desired power. The researcher should calculate the sample size with a biostatistician, based on the expected effect size and the variability.

### What is the difference between a main effect and an interaction?

A main effect is the effect of one factor averaged across the levels of the other factors. An interaction is the difference in the effect of one factor across the levels of another factor.

### How do I randomize the experimental units?

The researcher should use a random number generator to assign the experimental units to the treatment combinations. The randomization scheme should be recorded in the protocol.

### What should I do if the design becomes unbalanced?

If the design becomes unbalanced, the researcher should consult a biostatistician. The analysis should account for the imbalance, and the researcher should report the loss of experimental units.

### What is the role of the analysis plan in the template?

The analysis plan states the statistical model, the primary comparison, and the method for handling missing data. The plan should be written before the data are collected to prevent bias.

### How do I choose the levels of a factor?

The levels should be based on the biological question, the practical range, and the pilot data. The levels should be stated in the template and justified in the proposal.

### What reporting guidelines should I use?

The researcher should select the reporting guideline that is relevant to the study type from the EQUATOR Network. The guideline should be cited in the methods and the checklist should be completed.

## Using the Evidence

| Source | Best use in this topic | Important limitation |
|---|---|---|
| [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books) | official guidance | Check the linked page for current local requirements |
| [EQUATOR Network](https://www.equator-network.org/) | official guidance | Check the linked page for current local requirements |
| [Core Practices](https://publicationethics.org/core-practices) | official guidance | Check the linked page for current local requirements |

## Related Diagnostic Guides

- [Biosafety Level 4 Laboratories: Design, Operational Protocols, and Diagnostic Applications](/knowledge/diagnostics/emerging-tech/biosafety-level-4-laboratories-design-operational-protocols-and-diagnostic-applications)
- [Biosafety Protocols for Diagnostic Specimen Transport: Ensuring Safety and Compliance](/knowledge/diagnostics/emerging-tech/biosafety-protocols-for-diagnostic-specimen-transport-ensuring-safety-and-compliance)
- [Multiplex PCR: Design, Optimization, and Troubleshooting](/knowledge/diagnostics/molecular/multiplex-pcr-design-optimization-and-troubleshooting)

## 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.
- [Molecular detection of &lt,i&gt,Coxiella Burnetii&lt,/i&gt, in ovine abortions: evidence from a three-year surveillance in an endemic region.](https://doi.org/10.1016/j.vas.2025.100550). 2026.
- [Scenario design for infectious disease projections: Integrating concepts from decision analysis and experimental design.](https://doi.org/10.1016/j.epidem.2024.100775). 2024.
- [Tripartite binding mode of cohesin-dockerin complexes from Ruminococcus flavefaciens involving naturally truncated dockerins.](https://doi.org/10.1016/j.jbc.2025.110325). 2025.

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