Plackett-Burman Designs in Bioprocess Optimization

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

Quick Answer

• Plackett-Burman designs screen many bioprocess factors with few runs to identify which variables most influence an output before detailed optimization.
• Select factors from prior knowledge, assign high and low levels, then rank effects using half-normal plots or Pareto charts to choose critical variables.
• The design detects main effects only and cannot estimate interactions, curvature, or pure error without replication, so confirm findings with follow-up experiments.

At a Glance

Understanding Plackett-Burman Designs in Bioprocess Context

Plackett-Burman designs belong to a family of two-level fractional factorial designs used for factor screening. In bioprocess development, researchers often face a situation where many variables could influence a response, such as medium components, temperature, pH, agitation speed, inoculum size, or induction time. Testing each factor one at a time consumes resources and fails to capture how factors behave together. A Plackett-Burman design offers a structured way to examine many factors with relatively few experimental runs.

The design works by constructing a matrix where each factor is set at one of two levels, typically coded as -1 and +1. The number of runs is a multiple of four, and the design can accommodate up to one fewer factor than the number of runs. For example, a 12-run design can screen up to 11 factors, and a 20-run design can screen up to 19 factors. This efficiency makes the design attractive for early-stage bioprocess work where the goal is to narrow a long list of potential variables to a shorter list of critical ones.

The statistical foundation of the design rests on the assumption that the response can be approximated by a linear model with main effects. The design deliberately sacrifices the ability to estimate interactions and curvature to gain efficiency in the number of runs. This tradeoff is acceptable when the goal is screening, because the researcher wants to identify which factors deserve deeper investigation in a subsequent optimization study.

The design matrix is constructed from a specific pattern of plus and minus signs. The pattern is not arbitrary, and the arrangement determines which effects are confounded with each other. In a Plackett-Burman design, main effects are confounded with two-factor interactions, meaning that a large estimated effect for a factor could actually be caused by an interaction between two other factors. This is the central limitation of the design and the reason why results must be interpreted as screening outcomes instead of definitive estimates.

For bioprocess researchers, the practical value of the design lies in its economy. A study that would require dozens of runs to test each factor individually can be completed in a fraction of the time. This is especially useful when each run involves a fermentation, a culture, or an assay that takes days to complete. The design also provides a formal framework for deciding which factors to carry forward, which reduces the risk of overlooking an important variable.

The design is not appropriate for every situation. If you already know that interactions are important, or if you need precise estimates of factor effects, a full factorial or response surface design is a better choice. The Plackett-Burman design is a screening tool, and its results should always be followed by confirmation experiments.

Core Principles of Screening Designs

Main Effects and the Linear Model

The Plackett-Burman design assumes that the response can be approximated by a first-order model. In this model, the response is a function of the sum of the individual factor effects, with no interaction terms. The model is written as Y = b0 + b1X1 + b2X2 + ... + bkXk + error, where Y is the response, X values are the coded factor levels, and b values are the coefficients that estimate the effect of each factor.

The design estimates each main effect by comparing the average response at the high level of a factor with the average response at the low level. The difference between these averages, divided by two, gives the estimated effect. A factor with a large effect relative to the others is a candidate for further study.

The linear model assumption is important because it means the design cannot detect curvature. If the true relationship between a factor and the response is curved, the design will miss it. This is why the design is used for screening, not for optimization. Once you identify the critical factors, you move to a design that can model curvature, such as a central composite design or a Box-Behnken design.

Resolution and Confounding

The resolution of a design describes the pattern of confounding. A Plackett-Burman design is a resolution III design, which means that main effects are confounded with two-factor interactions. In practical terms, if you see a large effect for factor A, you cannot be certain whether the effect is truly due to A or whether it is due to an interaction between two other factors that happens to be confounded with A.

This confounding is the price you pay for the efficiency of the design. The design cannot distinguish between a main effect and a two-factor interaction, so you must treat the results as provisional. A factor that appears important in the screening should be tested in a follow-up experiment that can separate the main effect from the interaction.

The confounding pattern is determined by the specific arrangement of the design matrix. Different Plackett-Burman designs have different confounding patterns, and the pattern is documented in the design tables. When you analyze the results, you should be aware of which interactions are confounded with which main effects, because this information helps you interpret the results.

Efficiency and Run Count

The efficiency of the Plackett-Burman design comes from the fact that the number of runs is only slightly larger than the number of factors. A 12-run design can screen 12 factors, and a 20-run design can screen 19 factors. This efficiency is achieved by using a saturated design, where the number of runs equals the number of factors plus one.

The saturated design has no degrees of freedom left for estimating error. This means you cannot calculate a pure error estimate from the design alone. To estimate error, you need to add center points or replicate some of the runs. Center points are runs where all factors are set at the midpoint between the high and low levels. Replicates are runs where the same factor settings are repeated.

The choice of run count depends on the number of factors you need to screen. If you have 8 factors, a 12-run design gives you 4 extra degrees of freedom. If you have 15 factors, you need a 20-run design. The design should be chosen so that the number of runs is a multiple of four and is at least one more than the number of factors.

Planning a Plackett-Burman Experiment

Step 1: Define the Response and the Factor List

The first step is to define the response you want to optimize. In a bioprocess, the response could be product yield, cell density, enzyme activity, or any other measurable output. The response must be measurable, reproducible, and sensitive to changes in the factors.

The factor list should come from prior knowledge. Review the literature, your own preliminary experiments, and the experience of your laboratory team. The goal is to generate a list of factors that could plausibly affect the response. The list should be comprehensive, because the purpose of the design is to screen many factors at once. However, the list should also be realistic, because each factor requires a high and low level that you can actually set in the lab.

The number of factors determines the size of the design. If you have 12 factors, you need at least a 12-run design. If you have 19 factors, you need a 20-run design. The design can also include dummy factors, which are factors that are not actually varied but are included in the design matrix to estimate error. Dummy factors are assigned to columns in the design matrix that are not associated with a real factor.

Step 2: Setting High and Low Levels

Each factor must be assigned a high level and a low level. The levels should be chosen so that they represent a meaningful range for the factor. The range should be wide enough to produce a detectable effect, but not so wide that the high or low level is unrealistic or unsafe.

The levels should be based on prior data. If you have preliminary experiments, use the results to set the levels. If you are working from the literature, use the ranges reported in similar studies. The levels should be expressed in the units that you use in the lab, such as grams per liter, degrees Celsius, or pH units.

The choice of levels is critical because it determines the size of the effect you can detect. If the levels are too close together, the effect may be too small to detect. If the levels are too far apart, the factor may produce a response that is outside the range of your assay or that is not biologically meaningful.

Step 3: Constructing the Design Matrix

The design matrix is constructed using the Plackett-Burman pattern. The pattern is a sequence of plus and minus signs that is repeated and shifted to create the full matrix. The pattern for a 12-run design is different from the pattern for a 20-run design.

You can construct the matrix by hand, but it is easier to use statistical software. Most statistical packages include Plackett-Burman designs as a standard option. The software will generate the matrix, assign the factors to the columns, and provide the run order.

The run order should be randomized. Randomization is important because it spreads the effect of unknown variables over the entire experiment. If you run the experiment in a fixed order, you might confound the factor effects with a time trend, such as a drift in the instrument or a change in the environment. Randomization protects against this.

Step 4: Running the Experiment

The experiment is run according to the randomized order. Each run is a single experiment with the factor settings specified by the design matrix. The response is measured for each run, and the results are recorded.

The quality of the data depends on the care you take in running the experiment. The factor settings must be set accurately, and the response must be measured consistently. If a run fails, you need to decide whether to repeat it. A failed run can be repeated, but the repeat should be done at the same factor settings and in the same conditions.

The number of runs is small, so each run is valuable. You should plan the experiment so that you can complete all runs in a reasonable time. If the experiment takes several days, you should consider whether the conditions are stable over that time. If they are not, you may need to block the experiment or use a different design.

Analyzing Plackett-Burman Results

Estimating Factor Effects

The first step in the analysis is to estimate the effect of each factor. The effect is calculated as the difference between the average response at the high level and the average response at the low level. The effect is expressed in the units of the response.

The effects are then ranked in order of magnitude. The factors with the largest effects are the candidates for further study. The factors with small effects are likely to be unimportant, but you should not discard them without considering the confounding pattern.

The effects can be displayed in a Pareto plot, which shows the effects in descending order. The plot helps you see which factors stand out from the rest. The plot is a visual tool, and it should be used in combination with a statistical test.

Estimating Error and Significance

The design does not provide a pure error estimate, so you need to use one of several methods to assess significance. One method is to use the dummy factors. The dummy factors are columns in the design matrix that are not associated with a real factor. The effects of the dummy factors are estimates of the error, because they should be zero if the model is correct.

Another method is to use a normal probability plot of the effects. In a normal probability plot, the effects are plotted against the expected values from a normal distribution. The effects that are not significant will fall along a straight line, while the significant effects will deviate from the line.

A third method is to use a statistical test, such as the t-test or the F-test. The test compares the effect of each factor to the error estimate. The test requires an estimate of the error, which can come from the dummy factors or from replicated runs.

The choice of method depends on the design and the number of factors. If you have dummy factors, you can use them to estimate the error. If you do not have dummy factors, you can use the normal probability plot or the Pareto plot.

Interpreting the Results

The results of the analysis are a list of factors with their effects and significance. The factors that are significant are the critical factors that you should carry forward to the optimization stage. The factors that are not significant are the ones that you can drop from further consideration.

The interpretation should be cautious because of the confounding. A significant effect for a factor could be caused by an interaction between two other factors. You should not conclude that a factor is important until you have confirmed the result in a follow-up experiment.

The follow-up experiment can be a simple confirmation run, where you test the factor at the high and low levels while holding the other factors constant. Or it can be a more complex design, such as a full factorial design, that can separate the main effects from the interactions.

Practical Workflow for a Plackett-Burman Screening

Step 1: Define the Problem and the Response

Start by writing a clear statement of the problem. What is the response you want to improve? What is the current value of the response? What is the target value? The problem statement should be specific enough that you can measure the response and know when you have achieved the goal.

The response should be defined in terms of a measurable quantity. For example, the response could be the yield of a product in grams per liter, the activity of an enzyme in units per milliliter, or the growth of a culture in optical density units. The response should be measured with a method that is reliable and reproducible.

Step 2: Generate the Factor List

The factor list is the set of variables that you will test in the design. The list should be generated from a brainstorming session with your team, a review of the literature, and your own experience. The goal is to include all factors that could plausibly affect the response.

The list should be prioritized. You may have more factors than you can test in a single design. If you have more than 19 factors, you need to either reduce the list or use a larger design. The factors that are most likely to be important should be included in the design.

Step 3: Set the Levels

For each factor, you need to set a high level and a low level. The levels should be based on the literature, your preliminary experiments, or your experience. The levels should be realistic and safe.

The levels should be recorded in a table. The table should include the factor name, the units, the low level, and the high level. This table is the reference for the experiment.

Step 4: Construct the Design

Use statistical software to construct the design. The software will generate the design matrix, the run order, and the randomization. The design should be checked to ensure that it is a Plackett-Burman design and that the number of runs is appropriate for the number of factors.

The design matrix should be saved in a file. The file should include the run order, the factor settings, and a column for the response. The response column should be empty until you run the experiment.

Step 5: Run the Experiment

Run the experiment in the order specified by the design. For each run, set the factors to the specified levels, measure the response, and record the result. The response should be measured in the same way for each run.

The experiment should be run in a single session if possible. If the experiment takes multiple sessions, you should consider the effect of time on the response. If the response is sensitive to time, you may need to block the experiment or use a different design.

Step 6: Analyze the Results

Enter the results into the statistical software. The software will calculate the effects, the significance, and the ranking of the factors. The results should be reviewed and interpreted.

The results should be presented in a table. The table should include the factor name, the effect, the significance, and the ranking. The table should be used to decide which factors to carry forward.

Step 7: Confirm the Results

The results of the screening should be confirmed in a follow-up experiment. The follow-up experiment should test the significant factors at the high and low levels while holding the other factors constant. The follow-up experiment should be designed to confirm the direction and the magnitude of the effects.

The confirmation experiment is important because it protects you from the confounding in the screening design. If the confirmation experiment does not confirm the screening results, you need to investigate the discrepancy.

Records and Measurements

What to Record

The records for a Plackett-Burman experiment should include the design matrix, the run order, the factor settings, the response, and the date and time of each run. The records should also include any observations about the experiment, such as problems with the equipment, deviations from the protocol, or unexpected results.

The records should be kept in a laboratory notebook or an electronic system. The records should be complete enough that another researcher could reproduce the experiment. The records should also be stored in a way that is secure and accessible.

How to Measure the Response

The response should be measured using a method that is reliable and reproducible. The method should be validated before the experiment. The validation should include a test of the precision and the accuracy of the method.

The response should be measured in the same way for each run. The measurement should be done by the same person, using the same equipment, and at the same time of day. This reduces the variation in the measurement.

How to Handle Missing or Failed Runs

If a run fails, you need to decide whether to repeat it. A run can fail for many reasons, such as a contamination, a equipment failure, or a mistake in the factor settings. If the run fails, you should record the reason for the failure and decide whether to repeat the run.

If you repeat the run, you should use the same factor settings and the same conditions. The repeated run should be done as soon as possible. If you do not repeat the run, you need to analyze the data without the missing run. The missing run will reduce the degrees of freedom and may affect the results.

Common Failure Patterns and Troubleshooting

Failure Pattern 1: The Design Is Too Saturated

A saturated design has no degrees of freedom for error. If you do not add dummy factors or replicates, you cannot estimate the error. This means you cannot test the significance of the effects. The result is that you cannot distinguish between the real effects and the noise.

The solution is to add dummy factors or replicates to the design. Dummy factors are columns in the design matrix that are not associated with a real factor. The effects of the dummy factors are estimates of the error. Replicates are runs that are repeated at the same factor settings. The replicates provide a direct estimate of the error.

Failure Pattern 2: The Levels Are Too Close

If the levels are too close, the effect of the factor will be too small to detect. The result is that the factor will appear to be unimportant, even if it is actually important. The solution is to set the levels further apart. The levels should be set so that the effect is large enough to be detected.

Failure Pattern 3: The Confounding Is Not Considered

The confounding in the design means that a main effect could be caused by an interaction. If you do not consider the confounding, you may conclude that a factor is important when it is actually an interaction. The solution is to be aware of the confounding pattern and to confirm the results in a follow-up experiment.

Failure Pattern 4: The Run Order Is Not Randomized

If the run order is not randomized, the factor effects may be confounded with a time trend. For example, if the instrument drifts over time, the drift will be confounded with the factor effects. The solution is to randomize the run order.

Failure Pattern 5: The Response Is Not Measured Consistently

If the response is not measured consistently, the variation in the response will be larger than the variation in the factor effects. The result is that the effects will be less significant. The solution is to use a validated method and to measure the response in the same way for each run.

Limitations and Interpretation Boundaries

The Design Cannot Detect Interactions

The Plackett-Burman design is a main-effects design. It cannot detect interactions between factors. If the true relationship between the factors and the response includes interactions, the design will not detect them. The result is that the design may miss important factors or misidentify the factors.

The Design Cannot Detect Curvature

The design assumes a linear relationship between the factors and the response. If the true relationship is curved, the design will not detect the curvature. The result is that the design may miss an optimum that is inside the range of the factors.

The Design Is a Screening Tool

The design is a screening tool, not an optimization tool. The purpose of the design is to identify the factors that are important. The design cannot tell you the optimal level of the factors. To find the optimal level, you need to use a response surface design.

The Results Are Confounded

The results are confounded because the main effects are confounded with the two-factor interactions. The result is that you cannot be sure that a significant effect is due to the factor itself. The result should be confirmed in a follow-up experiment.

Safety and Regulatory Context

Biosafety Considerations

The Plackett-Burman design is a statistical tool, and it does not change the biosafety requirements of the experiment. The biosafety requirements are determined by the organism, the materials, and the procedures. You should follow the biosafety guidelines for your institution and for the organism you are working with.

Data Management and Sharing

The data from the experiment should be managed according to the data management and sharing policy of your institution and your funding agency. The NIH has a data management and sharing policy that requires researchers to plan for the management and sharing of data. The policy is described in the NIH data management and sharing policy. The plan should describe the data that will be collected, the methods for managing the data, and the plan for sharing the data.

Publication Ethics

The results of the experiment should be reported honestly and transparently. The publication should follow the reporting guidelines for the type of study. The EQUATOR Network provides a list of reporting guidelines for different types of studies. The guidelines help you report the methods and the results in a way that is clear and complete.

The publication should follow the core practices of the Committee on Publication Ethics. The core practices include the responsibilities of the authors, the peer review process, and the handling of misconduct. The authors should be the people who made a significant contribution to the work.

Researcher Identity

The researchers should have an ORCID identifier. The ORCID identifier is a unique identifier for researchers. The identifier is used to link the researcher to their publications and to their research activities. The ORCID identifier is described in the ORCID for researchers.

Professional Escalation Criteria

When to Seek Help

You should seek help if you are not sure how to construct the design, how to analyze the results, or how to interpret the results. You should also seek help if the results are not clear or if the results are not consistent with your expectations.

The help can come from a statistician, a colleague with experience in experimental design, or the statistical software support. The help should be sought before you run the experiment, not after.

When to Stop the Experiment

You should stop the experiment if you find a problem with the design, the materials, or the procedure. The problem could be a mistake in the design matrix, a problem with the equipment, or a problem with the materials. The problem should be fixed before you continue.

When to Escalate the Results

You should escalate the results if the results are unexpected or if the results have implications for the safety or the ethics of the work. The escalation should be to the appropriate person, such as the supervisor, the safety officer, or the ethics committee.

Decision Framework for Factor Retention and Follow-Up

Setting a Retention Threshold Before Analysis

The most common error in Plackett-Burman screening is deciding which factors to keep after looking at the results. This introduces bias because the researcher unconsciously adjusts the threshold to match expectations. Set the retention rule before you run the experiment and write it in the laboratory notebook. The rule should combine a statistical criterion with a practical criterion.

A practical retention rule uses the magnitude of the effect relative to the response mean. For example, you might decide that any factor with an effect smaller than 5 percent of the mean response will be dropped regardless of statistical significance. This rule prevents you from carrying forward a factor that is statistically significant but practically irrelevant. A fermentation yield of 12 grams per liter with a factor effect of 0.3 grams per liter is not worth optimizing.

The statistical criterion should be set from the error estimate. If you include dummy factors, the standard error of the effects can be calculated from the dummy factor effects. A factor is retained if its effect is larger than two times the standard error. This is a simple rule that works well for screening. The exact multiplier can be adjusted based on the number of dummy factors and the confidence level you want.

Write the retention rule in the protocol before the experiment starts. The rule should state the practical threshold, the statistical threshold, and the rule for combining them. A factor is retained if it passes either threshold, because a large effect that is not statistically significant may still be real but poorly estimated. A factor is dropped only if it fails both thresholds.

Using a Decision Matrix for Factor Disposition

After the analysis, assign each factor to one of three categories. The first category is factors to carry forward to optimization. These factors have large effects that pass both the practical and statistical thresholds. The second category is factors to confirm in a targeted experiment. These factors have moderate effects that pass one threshold but not the other. The third category is factors to drop. These factors have small effects that fail both thresholds.

The decision matrix is a table with the factor name, the effect size, the standard error, the t ratio, the practical threshold, and the disposition. The table forces a written decision for every factor. This record is valuable when you present the results to collaborators or reviewers because it shows that the factor selection was systematic and not arbitrary.

Factors in the confirm category deserve a small follow-up experiment. This experiment can be a two-factor or three-factor full factorial with center points. The purpose is to determine whether the moderate effect is real or an artifact of confounding. The follow-up experiment is smaller than the main screening and can be completed in a few days.

Factors in the drop category should not be revisited unless the confirm experiment produces unexpected results. If the confirm experiment shows that a dropped factor is important, the original screening may have had a level-setting problem. The levels may have been too close together to produce a detectable effect.

Handling Confounded Effects in the Decision

The confounding pattern of the design means that a large effect for one factor could be caused by an interaction between two other factors. The decision framework must account for this. When a factor has a large effect, check the confounding pattern for the design. The pattern is documented in the design tables or in the statistical software output.

If the factor is confounded with an interaction that is plausible based on the biology, treat the factor as a candidate for confirmation instead of a confirmed critical factor. For example, if temperature is confounded with the interaction between pH and agitation, and you know that pH and agitation can interact in your system, the temperature effect should be confirmed before you commit to optimizing temperature.

The confirmation experiment should be designed to separate the main effect from the interaction. A two-factor experiment with the confounded factors is sufficient. The experiment should include the high and low levels of both factors in a full factorial arrangement. This experiment will show whether the main effect is real or whether the interaction is responsible for the observed effect.

Recording the Decision Rationale

The decision matrix should be part of the permanent record of the experiment. The record should include the retention rule, the decision matrix, and the rationale for each disposition. The rationale should be written in a sentence or two for each factor. The record should be stored with the design matrix and the raw data.

The record is important for the publication of the results. The NIH data management and sharing policy requires that the data and the methods be documented in a way that allows others to understand and reproduce the work. The decision matrix is part of the methods because it shows how the factors were selected for the optimization stage.

The record also protects the researcher from criticism. A reviewer or a collaborator may question why a factor was dropped. The decision matrix provides a written answer that is based on the data and the pre-specified rule. This is more convincing than a verbal explanation that the factor was not important.

Common Decision Errors

The first common error is keeping too many factors. A screening design that retains half of the tested factors has not achieved the goal of reducing the factor list. The optimization stage will be larger and more expensive than necessary. The retention rule should be strict enough to reduce the list to a manageable number.

The second error is dropping a factor because the effect is not statistically significant when the design has no dummy factors. Without dummy factors, there is no error estimate and no statistical test. The effect should be evaluated on the practical criterion alone. A factor with a large effect should be retained even if the statistical significance cannot be calculated.

The third error is ignoring the direction of the effect. The sign of the effect tells you whether the high level or the low level produces the higher response. This information is used to set the direction of the optimization. A factor with a positive effect should be moved toward the high level in the optimization. A factor with a negative effect should be moved toward the low level.

The fourth error is failing to document the decision. The decision matrix is the record of the factor selection. Without the record, the selection cannot be audited or reproduced. The record should be completed at the time of the analysis, not after the optimization is finished.

Escalation Criteria for the Decision

The decision framework includes escalation criteria. If the screening results are ambiguous, the investigator should consult a statistician before proceeding. Ambiguity includes a case where many factors have similar effects and the retention rule does not clearly separate them. A statistician can help with the analysis and with the design of the confirm experiment.

If the screening results contradict prior knowledge, the investigator should stop and review the experiment. A factor that is known to be important from the literature but shows no effect in the screening may have levels that are too narrow. The levels should be checked against the literature values. The experiment may need to be repeated with wider levels.

If the screening results have safety implications, the investigator should escalate to the supervisor or the safety officer. A factor that shows a large effect on a hazardous response should be handled with caution. The optimization should not proceed until the safety implications are reviewed.

The decision framework is a practical tool that turns the statistical output into a management decision. The framework is applied after the analysis and before the optimization. The framework is documented in the experiment record and is used to justify the factor selection to collaborators, reviewers, and funding agencies.

Frequently Asked Questions

What is the difference between a Plackett-Burman design and a full factorial design?

A full factorial design tests all possible combinations of the factor levels. This allows you to estimate the main effects and the interactions. A Plackett-Burman design tests only a fraction of the combinations. This allows you to estimate the main effects, but not the interactions. The Plackett-Burman design is more efficient, but it provides less information.

How many factors can I screen in a Plackett-Burman design?

The number of factors is limited by the number of runs. A 12-run design can screen up to 12 factors. A 20-run design can screen up to 19 factors. The number of runs is a multiple of four, and the number of factors is one less than the number of runs.

How do I choose the high and low levels for each factor?

The levels should be based on the literature, your preliminary experiments, or your experience. The levels should be realistic and should represent the range of the factor that you are interested in. The levels should be wide enough to produce a detectable effect, but not so wide that they are unrealistic or unsafe.

How do I estimate the error in a Plackett-Burman design?

The design does not provide a pure error estimate. You can estimate the error by adding dummy factors or replicates to the design. Dummy factors are columns in the design matrix that are not associated with a real factor. Replicates are runs that are repeated at the same factor settings.

What is the confounding in a Plackett-Burman design?

The confounding is the pattern of the design. In a Plackett-Burman design, the main effects are confounded with the two-factor interactions. This means that a significant effect for a factor could be caused by an interaction between two other factors. The confounding is the price you pay for the efficiency of the design.

How do I confirm the results of a Plackett-Burman design?

The results should be confirmed in a follow-up experiment. The follow-up experiment should test the factors that are significant in the screening design. The follow-up experiment should be designed to confirm the results and to separate the main effects from the interactions.

What is the next step after a Plackett-Burman design?

The next step is to optimize the factors that are significant in the screening design. The optimization can be done using a response surface design, such as a central composite design or a Box-Behnken design. The response surface design will provide a model that can be used to find the optimal level of the factors.

What are the limitations of a Plackett-Burman design?

The design cannot detect interactions or curvature. The design is a screening tool, and it cannot be used for optimization. The results are confounded, so they must be confirmed in a follow-up experiment. The design is most useful when you have many factors and you need to identify the critical ones.

Using the Evidence

Related Diagnostic Guides

• Multiplex PCR: Design, Optimization, and Troubleshooting
• Method Validation in Clinical Laboratories: A Practical Protocol
• Animal Biology

References and Further Reading

• Research Methods Resources. National Library of Medicine.
• EQUATOR Network. EQUATOR Network.
• Core Practices. Committee on Publication Ethics.
• NIH Grants and Funding. National Institutes of Health.
• ORCID for Researchers. ORCID.
• Data Management and Sharing Policy. National Institutes of Health.
• Laboratory Biosafety Manual. World Health Organization.
• NCBI Literature Resources. National Center for Biotechnology Information.
• Experimental design methods for bioengineering applications.. Critical reviews in biotechnology, 2016.
• Enhancement of Biomass and Calcium Carbonate Biomineralization of Chlorella vulgaris through Plackett-Burman Screening and Box-Behnken Optimization Approach.. Molecules (Basel, Switzerland), 2020.
• Bioreactor process parameter screening utilizing a Plackett-Burman design for a model monoclonal antibody.. Journal of pharmaceutical sciences, 2015.

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