# Handling Missing Data in Contingency Tables ## Quick Answer - Missing observations in contingency tables can distort chi-square statistics and odds ratios, so the handling method must match the missingness mechanism and analysis goal. - Start with complete-case analysis as the baseline, then apply multiple imputation when missingness exceeds a small fraction and the data meet its assumptions. - No method recovers information lost to missingness, and every approach relies on assumptions that sensitivity analysis should test. ## Missing Data in Categorical Research Contingency tables are the workhorse of categorical data analysis in the life sciences. Researchers use them to examine associations between genotypes and phenotypes, treatment groups and outcomes, or exposure categories and disease status. The chi-square test of independence and the odds ratio are the standard tools for these tables. Both tools assume that the observed counts reflect the true population structure. When observations are missing, that assumption fails, and the resulting statistics can mislead. Missing data in categorical research arises from many sources. A laboratory assay fails for a subset of samples. A study participant does not return for a follow-up visit. A field observation is lost because a recorder malfunctioned. A histopathology slide is damaged during processing. Each of these events removes a cell from the table that would have contributed to a count. The question is not whether missing data affects the analysis, but how much and in which direction. The practical problem for a researcher is deciding what to do with incomplete tables. The options range from ignoring the problem to complex model-based approaches. Each option carries assumptions, and each can produce different conclusions from the same underlying data. This article walks through the decision process, the methods available, and the records and checks that make the analysis defensible. ## At a Glance | Missing Data Pattern | Recommended Approach | Key Assumption | Practical Limitation | | --- | --- | --- | --- | | Small fraction of cells missing, random across categories | Complete-case analysis | Missingness does not depend on the outcome or exposure | Loses power and may bias estimates if the assumption fails | | Moderate missingness, variables related to missingness are measured | Multiple imputation | Missingness is explained by observed variables | Requires careful model specification and diagnostic checking | | Large missingness or nonrandom missingness | Sensitivity analysis with a range of plausible scenarios | The true missingness mechanism lies within the tested range | Cannot prove the true answer, only bound the plausible range | ## The Structure of Missing Data in Contingency Tables A contingency table records counts of observations that fall into combinations of categories. A two-way table has rows for one variable and columns for another. Each cell contains the number of observations with that specific combination. The row totals and column totals summarize the marginal distributions. The chi-square test compares the observed cell counts to the counts expected if the two variables were independent. The odds ratio compares the odds of an outcome in one group to the odds in another group. Missing data enters this structure in two ways. Unit nonresponse removes an entire observation from the table. The observation has no values for any variable, so it contributes nothing to any cell. Item nonresponse removes only part of an observation. The observation has a value for one variable but not for another, so it cannot be placed in a cell even though it carries partial information. The distinction matters for analysis. Unit nonresponse reduces the sample size uniformly. Item nonresponse creates a table where some observations are partially classified. The researcher must decide whether to discard the partial observations or to use methods that incorporate them. ### Missingness Mechanisms The statistical literature classifies missing data by the relationship between missingness and the values of the variables. The classification determines which methods are valid. Missing completely at random means the probability of missingness is the same for all observations. A laboratory plate that is dropped and destroyed removes samples without regard to their values. This is the most benign form of missingness because the observed data is a random subsample of the full data. Missing at random means the probability of missingness depends on observed variables but not on the missing values themselves. For example, older samples may be more likely to fail an assay, and age is recorded for all samples. The missingness is explained by the observed data, so methods that condition on those variables can recover the missing information. Missing not at random means the probability of missingness depends on the missing value itself. A sample with an extreme measurement may be more likely to be lost, or a participant with a poor outcome may be more likely to drop out. This mechanism is the hardest to handle because the missingness carries information that the observed data cannot provide. The mechanism is not directly observable from the data. The researcher must reason from the study design and the data collection process. A record of why observations are missing is essential for this reasoning. The [National Library of Medicine research methods resources](https://www.ncbi.nlm.nih.gov/books) describe how the missingness mechanism shapes the choice of analytic approach. ## Complete-Case Analysis Complete-case analysis restricts the dataset to observations with no missing values. The contingency table is built only from these complete observations. This is the default approach in most statistical software and the first analysis that most researchers run. The method is valid when the missing data is missing completely at random. Under this mechanism, the complete cases are a random subsample of the full data, so the chi-square test and odds ratio are unbiased. The only cost is reduced sample size and reduced statistical power. The method becomes biased when the missing data is missing at random or missing not at random. If the probability of missingness depends on the outcome or on the exposure, the complete cases are not representative of the full population. The cell counts are distorted, and the chi-square test and odds ratio reflect the distorted distribution. ### When Complete-Case Analysis Is Acceptable Complete-case analysis is acceptable when the missing fraction is small and the researcher has reason to believe the missingness is random. A small missing fraction, such as a few percent of observations, is unlikely to change the conclusions of a chi-square test unless the test is borderline. The researcher should report the missing fraction and the reasoning for treating it as random. The method is also acceptable when the missingness is clearly mechanical. A failed assay, a lost sample, or a recording error removes observations without regard to their values. The researcher should document the cause of the missingness in the study records. ### The Limits of Complete-Case Analysis The limits appear when the missing fraction grows or when the missingness is related to the variables of interest. A study with 20 percent missing data loses a substantial portion of its information. The confidence intervals widen, and the power to detect a real association drops. The bias is more serious than the power loss. If the missingness depends on the outcome, the complete-case table shows a distorted relationship. A treatment that appears effective in the complete cases may be ineffective in the full data, or the reverse. The researcher cannot detect this distortion from the complete-case analysis alone. The [Committee on Publication Ethics core practices](https://publicationethics.org/core-practices) emphasize that data handling decisions should be transparent and reproducible. The complete-case analysis is the baseline against which other methods are compared, and the report should state the missing fraction and the reasons for the chosen approach. ## Multiple Imputation for Categorical Data Multiple imputation replaces each missing value with a set of plausible values. The analysis is run on each completed dataset, and the results are combined across the datasets. The method accounts for the uncertainty in the missing values by producing multiple versions of the completed data. The method is valid when the missing data is missing at random. The imputation model uses the observed variables to predict the missing values. If the observed variables contain the information needed to explain the missingness, the imputed values are unbiased. ### Building the Imputation Model The imputation model for categorical data must respect the categorical nature of the variables. A categorical variable with two categories can be imputed with logistic regression. A categorical variable with more than two categories requires a multinomial model. The model should include the variables that will be used in the analysis, because the imputation must preserve the relationships among those variables. The model should also include auxiliary variables that predict missingness. These are variables that are not part of the analysis but are related to the probability of missingness. Including them improves the imputation and reduces the bias. ### Combining the Results The analysis is run on each imputed dataset. The chi-square statistic and the odds ratio are computed for each dataset. The results are combined using the rules for multiple imputation, which average the estimates and adjust the variance for the between-dataset variability. The combined estimate is unbiased when the imputation model is correct. The variance is larger than the variance from a single complete dataset because it includes the uncertainty from the missing values. ### Practical Considerations Multiple imputation requires the researcher to specify the number of imputed datasets. The number should be large enough to produce stable estimates. The researcher should also check that the imputed values are plausible. An imputed value that falls outside the possible range for a category indicates a problem with the model. The [National Institutes of Health data management and sharing policy](https://sharing.nih.gov/data-management-and-sharing-policy) requires that research data be managed and shared in a way that supports reproducibility. The imputation model, the number of datasets, and the software code should be documented so that another researcher can reproduce the analysis. ## Sensitivity Analysis Sensitivity analysis tests how the conclusions change under different assumptions about the missing data. The researcher specifies a range of plausible missingness mechanisms and evaluates the analysis under each one. The result is a range of possible conclusions instead of a single point estimate. The method is essential when the missingness is missing not at random. No statistical method can recover the missing values when the missingness depends on the missing values themselves. The researcher can only bound the plausible range of the true answer. ### Constructing the Scenarios The scenarios should reflect the plausible mechanisms for the missingness. A scenario might assume that the missing observations are all in one category of the outcome. Another scenario might assume that the missing observations are distributed differently across the exposure groups. The scenarios should be informed by the study design and the reasons for the missingness. The analysis is run under each scenario. The chi-square statistic and the odds ratio are computed for each. The researcher reports the range of results and the conditions under which the conclusion changes. ### Interpreting the Results The sensitivity analysis does not produce a single answer. It produces a range of answers that correspond to the range of plausible assumptions. The researcher must judge whether the conclusion is robust across the range or whether it depends on the assumption. A conclusion that holds across all plausible scenarios is robust. A conclusion that changes under a plausible scenario is fragile. The report should state the fragility and the conditions under which the conclusion changes. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines that help researchers describe the sensitivity analysis in a transparent way. The guidelines emphasize that the assumptions and the scenarios should be stated explicitly. ## Practical Workflow for Missing Data The workflow for handling missing data in a contingency table follows a sequence of decisions. The researcher should document each decision and the reasoning behind it. ### Step 1: Assess the Missingness The first step is to describe the missing data. The researcher should compute the missing fraction for each variable and for each cell of the table. The pattern of missingness should be examined. Is the missingness concentrated in one variable or one category? Is it spread across the table? The researcher should also record the reasons for missingness. The data collection records should indicate whether the missingness was mechanical, such as a failed assay, or related to the study variables, such as a participant dropping out. ### Step 2: Determine the Missingness Mechanism The mechanism is determined by reasoning from the data collection process. The researcher should ask whether the missingness depends on the observed variables, the missing values, or neither. The answer determines the valid methods. The reasoning should be documented. The report should state the assumed mechanism and the evidence for that assumption. ### Step 3: Run the Complete-Case Analysis The complete-case analysis is the baseline. The chi-square test and the odds ratio are computed from the complete cases. The missing fraction is reported. ### Step 4: Apply Multiple Imputation If the missingness is missing at random and the missing fraction is moderate, multiple imputation is the next step. The imputation model is built, the analysis is run on each imputed dataset, and the results are combined. ### Step 5: Run the Sensitivity Analysis The sensitivity analysis tests the robustness of the conclusions. The scenarios are specified, the analysis is run under each, and the range of results is reported. ### Step 6: Report the Results The report should state the missing fraction, the missingness mechanism, the methods used, and the results of the sensitivity analysis. The report should follow the relevant reporting guideline from the [EQUATOR Network](https://www.equator-network.org/). ## Records and Measurements The records for missing data analysis are the foundation of the reproducibility. The researcher should maintain a data dictionary that describes each variable and the coding of missing values. The dictionary should state whether a missing value is a true absence or a failed measurement. The data collection records should indicate the reason for each missing observation. The reason should be recorded at the time of the missingness, not reconstructed later. A laboratory log that records failed assays and lost samples is the primary source for this information. The analysis records should include the software code for the complete-case analysis, the imputation model, and the sensitivity analysis. The code should be version-controlled and stored with the data. The [NIH data management and sharing policy](https://sharing.nih.gov/data-management-and-sharing-policy) describes the expectations for data management and sharing in NIH-funded research. ### The Missing Data Log A missing data log is a table that records each missing observation. The log should include the observation identifier, the variable with the missing value, the reason for the missingness, and the date of the missingness. The log is the primary evidence for the missingness mechanism. The log should be maintained throughout the data collection period. The researcher should not wait until the analysis stage to reconstruct the reasons for missingness. The reasons are most accurate when recorded at the time of the event. ### The Analysis Record The analysis record documents the decisions made during the analysis. The record should state the missing fraction, the assumed mechanism, the methods used, and the results. The record should also state the scenarios used in the sensitivity analysis and the conditions under which the conclusion changes. The analysis record is the basis for the methods section of the publication. The [Committee on Publication Ethics core practices](https://publicationethics.org/core-practices) require that the data handling be described accurately and transparently. ## Common Failure Patterns The following failure patterns appear in missing data analysis. The researcher should recognize them and take corrective action. ### Ignoring the Missing Data The most common failure is to ignore the missing data entirely. The researcher runs the analysis on the complete cases without reporting the missing fraction or the reasons for missingness. The result is a biased analysis that cannot be evaluated by the reader. The corrective action is to report the missing fraction and the missingness mechanism. The report should state the reasons for the missingness and the methods used to handle it. ### Treating Missing as a Category A common error is to treat missing values as a separate category in the contingency table. The missing values are counted as a category of the variable, and the chi-square test is run on the expanded table. This approach is valid only when the missingness is a meaningful category, such as a test that was not performed. When the missingness is a failed measurement, treating it as a category distorts the analysis. The missing category is not a biological category, and the chi-square test will reflect the missingness instead of the relationship between the variables. ### Using the Wrong Imputation Model The imputation model must match the categorical nature of the variable. A linear model for a categorical variable can produce imputed values that are not in the category set. The imputed values are then rounded to the nearest category, which introduces bias. The corrective action is to use a model that respects the categorical nature of the variable. Logistic regression for binary variables and multinomial models for categorical variables are the appropriate choices. ### Overlooking the Missingness Mechanism The missingness mechanism determines the valid methods. A researcher who assumes missing completely at random when the missingness is missing at random will produce biased results. The corrective action is to reason about the mechanism from the data collection process and to document the reasoning. ### Failing to Run the Sensitivity Analysis The sensitivity analysis is the only way to test the robustness of the conclusions to the missingness assumptions. A researcher who skips the sensitivity analysis cannot know whether the conclusions depend on the assumptions. The corrective action is to run the sensitivity analysis and to report the range of results. ## Welfare and Safety Context The handling of missing data has implications for the welfare of research subjects and the safety of the conclusions. A biased analysis can lead to a conclusion that a treatment is effective when it is not, or that a risk factor is harmless when it is not. The bias can affect decisions about patient care, public health, and environmental policy. The researcher has a responsibility to the research subjects to report the analysis accurately. The [Committee on Publication Ethics core practices](https://publicationethics.org/core-practices) state that researchers should report their methods and results honestly. The missing data handling is part of the methods, and the report should be transparent. The safety context applies to the conclusions. A conclusion that is based on a biased analysis is unsafe to act on. The sensitivity analysis is the tool that tests the safety of the conclusion. A conclusion that holds across the plausible range of missingness assumptions is safe to act on. A conclusion that changes under a plausible assumption is not. ## Professional Escalation Criteria The researcher should seek professional help when the missing data analysis exceeds their expertise. The following situations warrant escalation. ### Complex Missingness Patterns A missingness pattern that involves multiple variables and multiple mechanisms is beyond the scope of a simple analysis. The researcher should consult a biostatistician who has experience with missing data methods. ### Missing Not at Random When the missingness is missing not at random, the analysis requires specialized methods. The researcher should consult a biostatistician who can design the sensitivity analysis and interpret the results. ### Regulatory or Publication Scrutiny When the analysis is part of a regulatory submission or a publication that will be scrutinized, the researcher should have the analysis reviewed by a biostatistician. The review should cover the missingness mechanism, the methods, and the sensitivity analysis. ### The Analysis Is the Primary Outcome When the contingency table analysis is the primary outcome of the study, the missing data handling is a critical decision. The researcher should consult a biostatistician before the analysis is finalized. ## The Role of Reporting Guidelines Reporting guidelines provide a structure for describing the missing data analysis. The [EQUATOR Network](https://www.equator-network.org/) is the gateway to the reporting guidelines for different study types. The guidelines specify the information that should be reported about the missing data, the methods, and the sensitivity analysis. The researcher should select the reporting guideline that matches the study type. The guideline should be consulted at the design stage, not the publication stage. The guideline informs the data collection and the analysis decisions. The report should state the missing fraction, the missingness mechanism, the methods used, and the results of the sensitivity analysis. The report should also state the limitations of the analysis and the conditions under which the conclusions might change. ## The Role of Data Management The data management plan is the foundation of the missing data analysis. The plan should describe the data collection, the quality checks, and the handling of missing values. The [NIH data management and sharing policy](https://sharing.nih.gov/data-management-and-sharing-policy) requires that the plan be described in the grant application. The plan should specify the coding of missing values. A missing value should be coded in a way that distinguishes a genuine missing from a failed measurement. The plan should also specify the records that will be maintained for the missing data. The plan should be updated when the data collection reveals unexpected missingness. The update should be documented and the reasons for the update should be stated. ## The Role of the Researcher Identity The researcher identity is part of the reproducibility of the analysis. The [ORCID for Researchers](https://info.orcid.org/researchers) provides a persistent identifier that links the researcher to their work. The identifier should be used in the publication and in the data repository. The identifier supports the reproducibility of the analysis by making it possible to find the researcher and the data. The researcher should maintain their ORCID record and link it to their publications and datasets. ## The Grant and Funding Context The missing data analysis is part of the research that is funded by grants. The [NIH Grants and Funding](https://grants.nih.gov/) pages describe the expectations for the conduct of the research. The data management and sharing policy is part of the grant requirements. The grant application should describe the missing data handling in the data management plan. The plan should state the methods that will be used and the records that will be maintained. The plan should be realistic and should match the resources of the research team. The grant review will evaluate the data management plan. A plan that does not address the missing data handling is a weakness in the application. ## The Analysis of the Odds Ratio The odds ratio is the measure of association in a contingency table. The odds ratio is the ratio of the odds of the outcome in the exposed group to the odds in the unexposed group. The odds ratio is computed from the cell counts of the table. Missing data affects the odds ratio in the same way that it affects the chi-square test. The cell counts are biased, and the odds ratio reflects the bias. The complete-case analysis produces an odds ratio that is unbiased only when the missingness is completely random. The multiple imputation produces an odds ratio that is unbiased when the missingness is missing at random. The imputation model must include the variables that explain the missingness. The combined odds ratio is the average of the odds ratios from the imputed datasets. The sensitivity analysis produces a range of odds ratios that correspond to the range of missingness assumptions. The range is the evidence for the robustness of the conclusion. ### The Confidence Interval The confidence interval for the odds ratio is the range of values that are plausible for the true odds ratio. The confidence interval is computed from the variance of the odds ratio. The variance is affected by the missing data. The complete-case analysis produces a confidence interval that is too narrow when the missingness is not completely random. The interval does not reflect the uncertainty from the missingness. The multiple imputation produces a confidence interval that is wider because it includes the uncertainty from the missingness. The sensitivity analysis produces a range of confidence intervals that reflect the range of assumptions. ## The Chi-Square Test The chi-square test is the standard test of independence in a contingency table. The test compares the observed cell counts to the expected counts under independence. The test statistic is the sum of the squared differences between the observed and expected counts, divided by the expected counts. The chi-square test is affected by the missing data in two ways. The missing data reduces the sample size, which reduces the power of the test. The missing data also biases the cell counts, which biases the test statistic. The complete-case analysis is valid when the missingness is completely random. The test statistic is unbiased, and the p-value is valid. The multiple imputation is valid when the missingness is missing at random. The test statistic is combined across the imputed datasets, and the p-value reflects the uncertainty from the missingness. The sensitivity analysis produces a range of p-values for the range of missingness assumptions. The range is the evidence for the robustness of the conclusion. ### The Expected Counts The expected counts are the counts that would be expected if the two variables were independent. The expected count for a cell is the product of the row total and the column total divided by the total sample size. The expected counts are affected by the missing data because the row totals and column totals are affected. The complete-case analysis computes the expected counts from the complete cases. The expected counts are unbiased when the missingness is completely random. The multiple imputation computes the expected counts from the imputed datasets. The expected counts are unbiased when the missingness is missing at random. ## The Data Collection The data collection is the stage at which the missing data can be prevented. The data collection plan should include the checks that identify missing values at the time of collection. The plan should also include the records that capture the reasons for the missingness. The data collection should be designed to minimize the missingness. The design should include the checks that identify the missing values. The design should also include the procedures for the missing values. The data collection should be monitored for the missingness. The monitoring should identify the patterns of missingness and the causes. The monitoring should be documented and the findings should be used to improve the data collection. ### The Data Checks The data checks are the procedures that identify the missing values. The checks should be run at the time of the data collection. The checks should identify the missing values and the reasons for the missingness. The checks should be documented in the data management plan. The checks should be run on a regular basis. The findings should be recorded and the corrective actions should be taken. ## The Analysis of the Missing The analysis of the missing data is the process of the missingness and the methods. The analysis should be documented in the data management plan. The analysis should be the basis for the methods section of the publication. The analysis should be the following steps. The missing fraction should be computed. The missingness mechanism should be assessed. The complete-case analysis should be run. The multiple imputation should be run. The sensitivity analysis should be run. The results should be reported. The analysis should be the responsibility of the researcher. The researcher should have the expertise to the analysis. The researcher should seek the help of a biostatistician when the analysis is beyond their expertise. ## The Reporting of the Missing The reporting of the missing data is the process of the missingness and the methods. The report should be the following information: The missing fraction for each variable. The missingness mechanism. The methods used. The results of the sensitivity analysis. The conditions under which the conclusions change. The report should follow the reporting guidelines for the study type. The guidelines are available from the [EQUATOR Network](https://www.equator-network.org/). The report should be transparent and the methods should be reproducible. ## The Limitations of the Methods The methods for the missing data have limitations. The complete-case analysis is valid only when the missingness is completely random. The multiple imputation is valid only when the missingness is missing at random. The sensitivity analysis is valid only when the scenarios are plausible. The limitations should be stated in the report. The report should state the conditions under which the methods are valid. The report should also state the conditions under which the methods are not valid. The limitations should be the basis for the interpretation of the results. The results should be interpreted in the context of the limitations. The conclusions should be stated with the limitations. ## The Professional Escalation The professional escalation is the process of seeking the help of a biostatistician. The escalation should be the following situations: The missingness is missing not at random. The missingness pattern is complex. The analysis is the primary finding. The analysis is the subject of the regulatory review. The escalation should be the process of the consultation. The consultation should be the review of the missingness mechanism, the methods, and the sensitivity analysis. The consultation should be the documentation of the findings. ## Frequently Asked Questions ### What is the first step when I find missing data in my contingency table? Compute the missing fraction for each variable and each cell of the table. Record the reasons for the missingness from the data collection records. This assessment determines which methods are valid. ### When is complete-case analysis acceptable? Complete-case analysis is acceptable when the missing fraction is small and the missingness is completely random. The complete cases are a random subsample of the full data, so the chi-square test and the odds ratio are unbiased. ### What is the difference between missing at random and missing completely at random? Missing completely at random means the missingness is independent of all variables. Missing at random means the missingness depends on the observed variables but not on the missing values. The distinction determines whether multiple imputation is valid. ### How many imputed datasets should I create? The number should be large enough to produce stable estimates. The imputation model should be examined for the plausibility of the imputed values. The number should be documented in the report. ### What is the sensitivity analysis for missing data? The sensitivity analysis tests the conclusions under a range of plausible missingness assumptions. The analysis is run under each scenario, and the range of results is reported. The range is the evidence for the robustness of the conclusions. ### Can I treat missing values as a separate category in the table? Treating missing values as a category is valid only when the missingness is a meaningful category, such as a test that was not performed. When the missingness is a failed measurement, treating it as a category distorts the analysis. ### What should I report about the missing data in my publication? Report the missing fraction for each variable, the missingness mechanism, the methods used, and the results of the sensitivity analysis. Follow the reporting guidelines for the study type from the [EQUATOR Network](https://www.equator-network.org/). ### When should I consult a biostatistician? Consult a biostatistician when the missingness is missing not at random, when the missingness pattern is complex, or when the analysis is the primary finding of the study. The biostatistician can the sensitivity analysis and interpret the results. ## 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 Bioinformatics Guides - [Metabolomics Data Analysis Workflow: From Raw Data to Biological Insight](/knowledge/bioinformatics/metabolomics-data-analysis-workflow-from-raw-data-to-biological-insight) - [Metagenomics Data Analysis: From Raw Reads to Biological Insights](/knowledge/bioinformatics/metagenomics-data-analysis-from-raw-reads-to-biological-insights) - [Proteomics Data Analysis Workflow: From Raw Spectra to Biological Insights](/knowledge/bioinformatics/proteomics-data-analysis-workflow-from-raw-spectra-to-biological-insights) - [Spatial Omics Data Analysis: From Image Processing to Biological Interpretation](/knowledge/bioinformatics/spatial-omics-data-analysis-from-image-processing-to-biological-interpretation) - [Single-Cell Sequencing Analysis Pipeline: From Raw Data to Biological Insights](/knowledge/bioinformatics/single-cell-sequencing-analysis-pipeline-from-raw-data-to-biological-insights) ## 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. - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Primer, Pipelines, Parameters: Issues in 16S rRNA Gene Sequencing.](https://pubmed.ncbi.nlm.nih.gov/33627512). mSphere, 2021. - [Effects of missing data and data type on phylotranscriptomic analysis of stony corals (Cnidaria: Anthozoa: Scleractinia).](https://pubmed.ncbi.nlm.nih.gov/30677508). Molecular phylogenetics and evolution, 2019. - [Evidence of absence treated as absence of evidence: The effects of variation in the number and distribution of gaps treated as missing data on the results of standard maximum likelihood analysis.](https://pubmed.ncbi.nlm.nih.gov/32971285). Molecular phylogenetics and evolution, 2021. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.