# Reporting Longitudinal Data Analysis in Scientific Papers ## Quick Answer - Report longitudinal analyses by stating the model family, covariance structure, fixed and random effects, and estimation method in the methods section. - Describe how missing data were handled, which software was used, and how assumptions were verified before presenting results. - A key limitation is that no single reporting template fits every study design, so authors must adapt guidance to their specific data structure. ## Why Longitudinal Reporting Demands Specificity Longitudinal studies collect repeated measurements from the same subjects over time. This design creates correlated observations that violate the independence assumption of standard statistical tests. When authors report such analyses, they must explain how they accounted for this correlation, or readers cannot evaluate whether the chosen methods were appropriate. The challenge is that biological data often arrive with irregular timing, missing visits, and nonlinear trajectories. A plant growth experiment may record height weekly for some plants and monthly for others. A clinical study may lose participants between follow-up visits. An ecological survey may sample different numbers of individuals at each time point. Each of these situations requires a different modeling approach, and the manuscript must make those choices transparent. The [National Library of Medicine](https://www.ncbi.nlm.nih.gov/books) hosts authoritative biomedical texts that describe the statistical foundations of repeated-measures analysis. These resources help authors understand the assumptions behind different models and the language used to describe them. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines that help authors decide which details to include in their manuscripts. ## Core Principles for Reporting Longitudinal Methods ### State the Research Question and Design The methods section must begin with a clear statement of the study design. This includes the number of measurement occasions, the spacing between measurements, and whether the design is balanced or unbalanced. A balanced design has the same number of measurements at the same times for all subjects. An unbalanced design has varying numbers or timing of measurements. Authors should describe the primary outcome variable and how it was measured at each time point. They should also state whether the analysis addresses a fixed set of time points or models time as a continuous variable. This distinction affects the choice of statistical model and the interpretation of results. ### Specify the Statistical Model The model specification is the most important part of the methods section. Authors must state the type of model used, such as a linear mixed model, generalized estimating equation, or repeated-measures analysis of variance. Each model has different assumptions and produces different interpretations. For mixed models, the manuscript must describe: - The fixed effects, which are the predictors of interest such as treatment group, time, and their interaction - The random effects, which account for subject-to-subject variation - The covariance structure, which describes how measurements within a subject are correlated over time The covariance structure is a critical choice. Common options include compound symmetry, which assumes equal correlation between all time points, and autoregressive structures, which assume that measurements closer in time are more highly correlated. The choice of covariance structure affects the standard errors and the conclusions drawn from the data. ### Describe the Estimation Method Authors should state how the model parameters were estimated. Maximum likelihood and restricted maximum likelihood are the most common methods for mixed models. The choice between them affects how variance components are estimated and how models are compared. The estimation method also determines how missing data are handled. Some methods assume missing data are missing at random, meaning the probability of missingness depends on observed data. Other methods require stronger assumptions. The manuscript must state the missing-data assumption and justify it based on the study design. ### Report Software and Version The methods section should name the statistical software and version used for the analysis. This includes the specific package or function used to fit the model. Different software implementations can produce slightly different results for the same model, so this information is essential for reproducibility. The [National Institutes of Health](https://grants.nih.gov/) emphasizes the importance of reproducibility in funded research. The [Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy) requires that data and analysis methods be described in ways that allow others to verify the results. Naming the software and version is a basic step toward this goal. ## Reporting Model Assumptions and Diagnostics ### Verify and Report Assumptions Every statistical model makes assumptions about the data. The methods section must state which assumptions were checked and how. For longitudinal models, the key assumptions include: - The residuals are normally distributed - The variance is constant across time and groups - The relationship between the outcome and predictors is linear - The covariance structure is correctly specified The authors should describe the diagnostic plots and tests used to verify these assumptions. This includes residual plots, quantile-quantile plots, and tests for heteroscedasticity. The results of these checks should be summarized in the results section or in supplementary materials. ### Report How Violations Were Handled When assumptions are violated, the authors must describe the corrective actions taken. This could include transforming the outcome variable, using a robust estimation method, or changing the covariance structure. The manuscript should state what was done and why. If the authors chose not to correct a violation, they should acknowledge this limitation. This transparency allows readers to judge the validity of the results and the strength of the conclusions. ## Reporting Results from Longitudinal Models ### Present Effect Sizes and Confidence Intervals The results section should report the estimated effects with their confidence intervals. A confidence interval provides a range of plausible values for the effect size, which is more informative than a p-value alone. The authors should report the estimated difference between groups at each time point or the overall effect of time. For mixed models, the results should include the estimated fixed effects, their standard errors, and the confidence intervals. The authors should also report the variance components, which describe the variation between subjects and within subjects over time. ### Report Model Comparison and Selection When multiple models were compared, the authors should describe the comparison process. This includes the criteria used, such as the Akaike information criterion or the Bayesian information criterion. The authors should state which model was selected and why. The comparison should be described in enough detail that a reader could reproduce the selection process. This includes the set of candidate models, the criteria used, and the final choice. ### Present Results in Tables and Figures Longitudinal results are often best presented graphically. A plot of the mean outcome over time for each group, with error bars or confidence bands, is a standard way to show the results. The figure should include the raw data points or a summary of the data at each time point. Tables should report the estimated effects, standard errors, and confidence intervals for each time point or for the overall effect. The table should also include the number of observations at each time point, especially if there is missing data. ## The At a Glance Table | Reporting Element | What to Include | Common Error | | --- | --- | --- | | Model specification | Fixed effects, random effects, covariance structure, estimation method | Omitting the covariance structure | | Missing data | Assumption made, method used, proportion of missing data | Ignoring missing data entirely | | Software | Program name, version, package or function | Naming only the program without the version | | Assumption checks | Diagnostic plots, tests, and how violations were handled | Claiming assumptions were met without evidence | | Effect sizes | Estimated effects, standard errors, confidence intervals | Reporting only p-values | | Model selection | Candidate models, comparison criteria, final choice | Describing the final model without the selection process | ## Practical Workflow for Reporting Longitudinal Analyses ### Step 1: Document the Data Structure Before writing the methods section, the authors should document the structure of the data. This includes the number of subjects, the number of time points, and the pattern of missing data. The authors should also record how the data was collected and any changes to the protocol during the study. This documentation should be kept in a study file that includes the raw data, the analysis code, and the output. The [ORCID for Researchers](https://info.orcid.org/researchers) page explains how researchers can maintain a record of their work and link it to their publications. This record helps establish the provenance of the analysis. ### Step 2: Write the Methods Section The methods section should be written in the order that the analysis was performed. Start with the data description, then the model specification, then the estimation method, and finally the assumptions and diagnostics. Each step should be described in enough detail that a reader could reproduce the analysis. The authors should use the reporting guidelines from the [EQUATOR Network](https://www.equator-network.org/) to check that all required elements are present. These guidelines are designed to improve the transparency and completeness of research reporting. ### Step 3: Write the Results Section The results section should present the findings in the same order as the methods. Start with the descriptive statistics, then the model results, then the diagnostics. The authors should report the effect sizes with confidence intervals and the variance components. The results should be presented in tables and figures that are clearly labeled and referenced in the text. The authors should also report the number of observations used in each analysis, since this can vary due to missing data. ### Step 4: Review for Completeness Before submission, the authors should review the methods and results sections against a checklist. This checklist should include all the elements described in this article. The authors should also ask a colleague to review the sections to ensure that the description is clear and complete. The [Committee on Publication Ethics](https://publicationethics.org/core-practices) describes the responsibilities of authors in reporting their work accurately. This includes ensuring that the methods are described in enough detail for others to evaluate the work. ## Common Failure Patterns in Longitudinal Reporting ### Failure to Describe the Covariance Structure Many manuscripts state that a mixed model was used but do not describe the covariance structure. This is a serious omission because the covariance structure affects the standard errors and the significance of the effects. The reader cannot evaluate the results without this information. The authors should state the covariance structure and the basis for choosing it. This could be based on the design of the study, the pattern of correlations in the data, or a comparison of models with different structures. ### Failure to Report Missing Data Missing data is common in longitudinal studies, but many manuscripts do not describe how it was handled. The authors should report the proportion of missing data, the pattern of missingness, and the method used to handle it. This includes the assumption about the missing data mechanism. The authors should also report the number of observations used in each analysis. This is important because the number of observations can vary between analyses if different variables have different amounts of missing data. ### Failure to Report the Software Version Many manuscripts state that the analysis was performed in a particular software but do not give the version. This makes it difficult for readers to reproduce the analysis, since different versions of the software may produce different results. The authors should report the software version and the specific package or function used. This information should be included in the methods section or in the supplementary materials. ### Failure to Report Diagnostics Some authors state that the assumptions were checked but do not describe the checks. This is not sufficient for a reader to evaluate the validity of the results. The authors should describe the diagnostic plots and tests used and summarize the findings. The authors should also describe how violations of assumptions were handled. If the assumptions were not met, the authors should state this and describe the implications for the results. ## Records and Measurements for Longitudinal Analyses ### Data Documentation The authors should maintain a data dictionary that describes each variable in the dataset. This includes the variable name, the type of data, the units, and the coding scheme. The data dictionary should also describe the time variable and how it was recorded. The data dictionary should be included in the supplementary materials or in a public repository. This allows readers to understand the data and to reproduce the analysis. ### Analysis Code The authors should keep the analysis code in a file that can be shared with the manuscript. The code should be commented to explain the steps and the decisions made. The code should also be versioned so that the authors can track changes over time. The [National Institutes of Health](https://sharing.nih.gov/data-management-and-sharing-policy) describes the expectations for sharing data and analysis methods. The authors should follow these expectations when preparing their manuscript. ### Output Files The authors should keep the output files from the analysis, including the model results and the diagnostic plots. These files should be stored with the data and the code so that the analysis can be reproduced. The output files should be organized in a way that makes it easy to find the results for each model. The authors should also keep a log of the analysis decisions, including the models that were considered and the reasons for the final choice. ## Quality Controls for Longitudinal Analyses ### Verify the Data Structure Before fitting any model, the authors should verify that the data is structured correctly. This includes checking that the subject identifiers are unique, that the time points are correct, and that the outcome variable is in the correct format. The authors should also check for duplicate records and for inconsistencies in the data. This includes checking that the values are within the expected range and that the time points are in the correct order. ### Verify the Model Specification The authors should verify that the model specification matches the design of the study. This includes checking that the fixed effects include the variables of interest and that the random effects capture the correct level of variation. The authors should also check that the covariance structure is appropriate for the data. This can be done by comparing the fit of models with different covariance structures and by examining the residuals. ### Verify the Results The authors should verify the results by running the analysis with a different software or by using a different method. This helps to confirm that the results are not an artifact of the software or the method. The authors should also check the results for consistency with the descriptive statistics. For example, the estimated effects should be consistent with the observed differences between groups at each time point. ## Limitations of Longitudinal Analyses ### Assumptions About Missing Data Longitudinal analyses often rely on assumptions about the missing data mechanism. The most common assumption is that the data is missing at random, which means that the probability of missingness depends on the observed data. If this assumption is not met, the results may be biased. The authors should state the assumption and the basis for it. They should also consider conducting a sensitivity analysis to assess the impact of the assumption on the results. ### Assumptions About the Covariance Structure The covariance structure is a simplification of the true correlation structure of the data. The choice of the covariance structure can affect the results, and the authors should consider the sensitivity of the results to this choice. The authors should compare models with different covariance structures and report the results of the comparison. This helps the reader to understand the impact of the choice on the results. ### Assumptions about the Functional Form The model assumes a particular relationship between the outcome and the predictors. This relationship is often linear, but the true relationship may be nonlinear. The authors should check the functional form and consider using a more flexible model if the relationship is nonlinear. The authors should also consider the possibility of interactions between the predictors and time. The model should include the interactions if they are of interest or if they are needed to describe the data. ## Safety and Regulatory Context ### Data Sharing Requirements The [National Institutes of Health](https://sharing.nih.gov/data-management-and-sharing-policy) requires that data and analysis methods be shared in a way that allows others to verify the results. This includes the data, the analysis code, and the documentation of the methods. The authors should be aware of these requirements and should plan for data sharing when they design the study. The authors should also be aware of the requirements of the journal and the funding agency. ### Publication Ethics The [Committee on Publication Ethics](https://publicationethics.org/core-practices) describes the responsibilities of authors in reporting their research. This includes the responsibility to report the methods accurately and to describe the limitations of the analysis. The authors should also be aware of the requirements for authorship and for the declaration of conflicts of interest. The authors should follow the policies of the journal and the publisher. ### Reproducibility The [National Institutes of Health](https://grants.nih.gov) emphasizes the importance of reproducibility in research. The authors should ensure that the analysis can be reproduced by others, which requires a complete description of the methods and the data. The authors should also ensure that the data is stored in a way that is accessible to others. This includes using a data repository and providing the data in a format that can be read by others. ## Professional Escalation Criteria ### When to Seek Statistical Advice The authors should seek statistical advice if they are uncertain about the model specification or the interpretation of the results. This is particularly important for complex analyses, such as those with nonlinear trajectories or with a large amount of missing data. The authors should also seek advice if the results are unexpected or if the diagnostics indicate problems with the model. A statistician can help to identify the cause of the problem and to suggest alternative approaches. ### When to Revise the Analysis The authors should revise the analysis if the diagnostics indicate that the model is not appropriate. This includes cases where the residuals are not normally distributed, where the variance is not constant, or where the covariance structure is not appropriate. The authors should also revise the analysis if the results are not consistent with the descriptive statistics or if the model does not converge. The authors should document the revisions and the reasons for the changes. ### When to Seek Peer Review The authors should seek peer review of the analysis before submission. This can be done by asking a colleague to review the methods and the results. The colleague should be able to reproduce the analysis and to identify any errors or omissions. The authors should also consider submitting the analysis to a preprint server or to a repository for the analysis. This allows others to review the analysis and to provide feedback. ## A Decision Framework for Selecting and Reporting Longitudinal Model Specifications Choosing the correct longitudinal model is the most consequential decision in the analysis, yet many manuscripts describe the model selection process poorly or omit it entirely. Reviewers and readers need to understand also which model was used but also why it was chosen over the alternatives. This section provides a practical decision framework that authors can apply before writing the methods section, ensuring that the model specification is defensible and that the reporting follows logically from the decisions made. ### The Three-Stage Model Selection Process The selection of a longitudinal model should proceed through three distinct stages, each with its own reporting requirements. The first stage is the structural decision, where the author determines whether the outcome is continuous, binary, count, or time-to-event. This decision determines the model family, such as linear mixed models for continuous outcomes, generalized linear mixed models for binary or count outcomes, and joint models for time-to-event outcomes with longitudinal predictors. The second stage is the correlation decision, where the author chooses how to model the within-subject correlation over time. This includes the choice between random effects, residual covariance structures, or both. The third stage is the estimation decision, where the author selects the estimation method and the approach to missing data. Each stage produces a decision that must be reported in the methods section. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines that help authors structure this information. The guidelines emphasize that the reporting should be complete enough for a reader to reproduce the analysis without guessing at any of the decisions. ### Stage One: Structural Decisions Based on Outcome Type The outcome variable type is the first and most important determinant of the model family. A continuous outcome measured on an interval or ratio scale is typically analyzed with a linear mixed model. A binary outcome, such as the presence or absence of a condition, requires a generalized linear mixed model with a logit or probit link. A count outcome, such as the number of events in a time interval, requires a generalized linear mixed model with a Poisson or negative binomial distribution. The author must state the outcome type explicitly in the methods section. This statement should include the measurement scale and the range of possible values. For example, a visual analog scale for pain is a continuous outcome that is bounded between zero and 100. A count of hospitalizations is a discrete outcome with a lower bound of zero and no upper bound. The choice of model family also depends on the distribution of the outcome. If the outcome is continuous but heavily skewed, the author may need to transform the outcome or use a generalized linear model with a non-normal distribution. The manuscript must describe the distributional assumptions and how they were verified. The [National Library of Medicine](https://www.ncbi.nlm.nih.gov/books) hosts texts that describe the statistical foundations of these model families. These resources help authors understand the assumptions behind each model and the language used to describe them in manuscripts. ### Stage 2: Correlation Decisions The correlation structure is the defining feature of longitudinal data. The author must decide how to model the fact that measurements from the same subject are more similar than measurements from different subjects. There are two main approaches: random effects and residual covariance structures. Random effects capture the between-subject variation. A random intercept allows each subject to have a different baseline level. A random slope allows each subject to have a different trajectory over time. The author must decide which random effects to include based on the study design and the research question. Residual covariance structures capture the within-subject correlation after accounting for the fixed and random effects. Common structures include compound symmetry, which assumes equal correlation between all time points, and autoregressive structures, which assume that measurements closer in time are more highly correlated. The author must choose the structure that best matches the pattern of correlation in the data. The decision between random effects and residual covariance structures is not always clear. Some models include both, such as a random intercept with an autoregressive residual structure. The author should compare models with different combinations and report the comparison in the manuscript. ### Stage 3: Estimation and Missing Data Decisions The estimation method determines how the model parameters are calculated from the data. Maximum likelihood and restricted maximum likelihood are the most common methods for linear mixed models. Generalized linear mixed models use approximations such as Laplace approximation or adaptive quadrature. The estimation method also determines how missing data are handled. The author must state the assumption about the missing data mechanism. The most common assumption is that the data are missing at random, meaning the probability of missingness depends on the observed data. The author must justify this assumption based on the study design and the pattern of missingness. The [National Institutes of Health](https://sharing.nih.gov/data-management-and-sharing-policy) requires that data and analysis methods be described in ways that allow others to verify the results. This includes the estimation method and the missing data assumption. The author should report the software and version used, as different implementations can produce different results. ### A Practical Decision Table for Model Selection The following table summarizes the decisions that must be made at each stage of the model selection process. The author should complete this table before writing the methods section and include the relevant information in the manuscript. | Decision Stage | Question to Answer | Reporting Requirement | | --- | --- | --- | | Outcome type | What is the scale of the outcome variable? | State the outcome type and distribution | | Model family | Which model family matches the outcome? | State the model family and link function | | Random effects | Which subject-level effects are included? | List the random effects and their structure | | Covariance structure | How is within-subject correlation modeled? | State the covariance structure and the basis for the choice | | Estimation method | How are the parameters estimated? | State the estimation method and software | | Missing data | What is the missing data assumption? | State the assumption and the handling method | | Model comparison | How was the final model selected? | Describe the candidate models and the criteria | The author should complete this table for each analysis in the manuscript. This ensures that the methods section is complete and that the reporting is consistent across all analyses. ## Troubleshooting Common Model Selection Problems ### When the Model Does Not Converge A common problem is that the model does not converge, meaning the estimation algorithm cannot find a stable solution. This often happens when the model is too complex for the data, such as when there are too many random effects or when the covariance structure is too flexible. The author should simplify the model by removing random effects or by using a simpler covariance structure. The author should also check for problems with the data, such as extreme outliers or collinearity among predictors. The manuscript should report the convergence problems and the steps taken to resolve them. ### When the Residuals Show a Pattern The residuals from the model should be randomly scattered around zero with no pattern. If the residuals show a pattern, such as a curve or a fan shape, the model is misspecified. The author should consider adding a nonlinear term for time, transforming the outcome, or changing the covariance structure. The manuscript should describe the diagnostic plots and the pattern observed. The author should also report the corrective action taken and the results of the revised model. ### When the Missing Data Are Not Missing at Random If the missing data mechanism is not missing at random, the standard methods will produce biased results. The author should consider a sensitivity analysis using a pattern-mixture model or a selection model. These models allow the missing data mechanism to depend on the unobserved values. The manuscript should report the sensitivity analysis and the impact of the missing data assumption on the results. This is important for the reader to judge the robustness of the conclusions. ## Records and Measurements for Model Selection ### The Model Selection Log The author should keep a log of all models considered during the analysis. This log should include the model specification, the estimation method, the convergence status, and the fit statistics. The log should also include the reasons for rejecting each model. This log is not necessarily included in the manuscript, but it is used to write the methods section. The author can refer to the log to describe the model selection process accurately. ### The Data Dictionary The data dictionary describes each variable in the dataset. This includes the variable name, the type of data, the units, and the coding scheme. The data dictionary should also describe the time variable and how it was recorded. The data dictionary should be included in the supplementary materials or in a public repository. This allows readers to understand the data and to reproduce the analysis. ### The Analysis Code The analysis code should be kept in a file that can be shared with the manuscript. The code should be commented to explain the steps and the decisions made. The code should also be versioned so that the author can track changes over time. The [National Institutes of Health](https://sharing.nih.gov/data-management-and-sharing-policy) describes the expectations for sharing data and analysis methods. The author should follow these expectations when preparing the manuscript. ## Common Failure Patterns in Model Selection Reporting ### Failure to Report the Model Selection Process Many manuscripts describe the final model but do not describe the process used to select it. This is a serious omission because the reader cannot evaluate whether the final model is appropriate. The author should describe the candidate models and the criteria used to select the final model. ### Failure to Report the Basis for the Covariance Structure The covariance structure is a critical choice that affects the standard errors and the significance of the effects. Many manuscripts state the covariance structure but do not describe the basis for the choice. The author should state whether the choice was based on the design, the data, or a comparison of models. ### Failure to Report the Missing Data Assumption The missing data assumption is a critical choice that affects the validity of the results. Many manuscripts do not state the assumption or the basis for it. The author should state the assumption and justify it based on the study design and the pattern of missingness. ### Failure to Report the Software Version The software version is important for reproducibility. Different versions of the software may produce different results for the same model. The author should report the software version and the specific package or function used. ## Quality Controls for Model Selection ### Verify the Model Specification The author should verify that the model specification matches the design of the study. This includes checking that the fixed effects include the variables of interest and that the random effects capture the correct level of variation. The author should also check that the covariance structure is appropriate for the data. This can be done by comparing the fit of models with different covariance structures and by examining the residuals. ### Verify the Results The author should verify the results by running the analysis with a different software or by using a different method. This helps to confirm that the results are not an artifact of the software or the method. The author should also check the results for consistency with the descriptive statistics. For example, the estimated effects should be consistent with the observed differences between groups at each time point. ### Verify the Reproducibility The author should ensure that the analysis can be reproduced by others. This requires a complete description of the methods and the data. The author should also ensure that the data is stored in a way that is accessible to others. The [Committee on Publication Ethics](https://publicationethics.org/core-practices) describes the responsibilities of authors in reporting their work accurately. This includes ensuring that the methods are described in enough detail for others to evaluate the work. ## Professional Escalation Criteria ### When to Seek Statistical Advice The author should seek statistical advice if they are uncertain about the model specification or the interpretation of the results. This is particularly important for complex analyses, such as those with nonlinear trajectories or with a large amount of missing data. The author should also seek advice if the results are unexpected or if the diagnostics indicate problems with the model. A statistician can help to identify the cause of the problem and to suggest alternative approaches. ### When to Revise the Analysis The author should revise the analysis if the diagnostics indicate that the model is not appropriate. This includes cases where the residuals are not normally distributed, where the variance is not constant, or where the covariance structure is not appropriate. The author should also revise the analysis if the results are not consistent with the descriptive statistics or if the model does not converge. The author should document the revisions and the reasons for the changes. ### When to Seek Peer Review The author should seek peer review of the analysis before submission. This can be done by asking a colleague to review the methods and the results. The colleague should be able to reproduce the analysis and to identify any errors or omissions. The author should also consider submitting the analysis to a preprint server or to a repository for the analysis. This allows others to review the analysis and to provide feedback. ## Frequently Asked Questions ### What is the difference between a mixed model and a repeated-measures ANOVA? A repeated-measures ANOVA is a special case of a mixed model that assumes a compound covariance structure and requires complete data. A mixed model is more flexible and can handle unbalanced data and different covariance structures. The mixed model is generally preferred for longitudinal data because it can handle missing data and does not require the assumption of compound covariance. ### How do I choose the covariance structure for my longitudinal model? The covariance structure should be chosen based on the design of the study and the pattern of correlations in the data. The authors should compare models with different covariance structures and select the one that fits the data best. The comparison should be based on the likelihood of the model or on the information criteria. ### What should I do if my data has missing values? The authors should describe the pattern of missing data and the assumption about the missing data mechanism. The analysis should be conducted under the assumption that the data is missing at random, and the authors should consider a sensitivity analysis to assess the impact of the assumption. The authors should also report the proportion of missing data and the number of observations used in each analysis. ### How do I report the results of a longitudinal analysis? The results should include the estimated effects with confidence intervals, the variance components, and the results of the diagnostics. The results should be presented in tables and figures that are labeled and referenced in the text. The authors should also report the number of observations used in each analysis. ### What software should I use for longitudinal analysis? The choice of software depends on the analysis and the preferences of the authors. Common options include R, SAS, Stata, and SPSS. The authors should report the software and version used in the analysis. ### How do I describe the model in the methods section? The methods section should describe the model specification, including the fixed effects, the random effects, the covariance structure, and the estimation method. The authors should also describe the assumptions and the diagnostics used to verify them. ### What are the common errors in reporting longitudinal analyses? Common errors include omitting the covariance structure, failing to report missing data, failing to report the software version, and failing to report the diagnostics. The authors should use a checklist to ensure that all the elements are present. ### How do I ensure that my analysis is reproducible? The authors should keep the data, the analysis code, and the output files in a repository. The methods should be described in enough detail that a reader can reproduce the analysis. The authors should also report the software and version used. ## Related Bioinformatics Guides - [Longitudinal Microbiome Data Analysis: Methods and Best Practices](/knowledge/bioinformatics/longitudinal-microbiome-data-analysis-methods-and-best-practices) - [Metabolomics Data Analysis in R: A Practical Workflow](/knowledge/bioinformatics/metabolomics-data-analysis-in-r-a-practical-workflow) - [Microbiome Data Analysis in R: A Practical Guide for Compositional Data](/knowledge/bioinformatics/microbiome-data-analysis-in-r-a-practical-guide-for-compositional-data) - [Genomic Data Analysis Tools: A Comparative Guide for Researchers](/knowledge/bioinformatics/genomic-data-analysis-tools-a-comparative-guide-for-researchers) - [Lipidomic Analysis: A Beginner's Guide to Workflows and Data Interpretation](/knowledge/bioinformatics/lipidomic-analysis-a-beginner-s-guide-to-workflows-and-data-interpretation) ## 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. - [Tooth wear and bruxism: A scoping review.](https://pubmed.ncbi.nlm.nih.gov/38574847). Journal of dentistry, 2024. - [Survival in a Real-World Cohort of Patients With Transthyretin Amyloid Cardiomyopathy Treated With Tafamidis: An Analysis From the Transthyretin Amyloidosis Outcomes Survey (THAOS).](https://pubmed.ncbi.nlm.nih.gov/38909877). Journal of cardiac failure, 2025. - [Initial data analysis for longitudinal studies to build a solid foundation for reproducible analysis.](https://pubmed.ncbi.nlm.nih.gov/38809844). PloS one, 2024. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.