# How to Choose the Right Statistical Test for Your Biological Data ## Key Takeaways - **Data Type and Group Structure Dictate Test Choice:** The fundamental step in selecting a statistical test involves classifying biological data as continuous (e.g., gene expression levels, growth rates), ordinal (e.g., disease severity scores), or categorical (e.g., presence/absence of a variant). Subsequently, one must determine the number of groups being compared and whether observations are paired (e.g., pre- and post-treatment measurements on the same subject) or independent. - **Assumption Verification is Critical for Parametric Tests:** Parametric statistical tests, such as Student's t-test and ANOVA, rely on assumptions like normality of data distribution and homogeneity of variances. Violations of these assumptions, verifiable with tests like Shapiro-Wilk for normality and Levene's for variance, necessitate the use of nonparametric alternatives (e.g., Mann-Whitney U, Kruskal-Wallis) or data transformations to avoid misleading results. - **Categorical Data Requires Frequency-Based Analysis:** For categorical biological data (e.g., survival counts, genotype frequencies), tests like the Chi-square test of independence or Fisher's exact test (for small sample sizes) are employed to compare observed frequencies against expected frequencies, assessing associations between variables. McNemar's test is specifically for paired categorical data. - **Relationships Between Continuous Variables Utilize Correlation and Regression:** When investigating associations between continuous biological variables (e.g., protein concentration and cell viability), Pearson correlation is used for linear relationships, while Spearman rank correlation is suitable for monotonic relationships. Simple and multiple linear regression models predict outcomes based on one or more explanatory variables, respectively. - **Documentation of Analytical Decisions is Paramount for Reproducibility:** A detailed audit trail documenting data structure, assumption checks (including specific test statistics and p-values), rationale for test selection, and any data transformations is essential for ensuring the reproducibility and transparency of biological data analysis, as mandated by policies like the NIH Data Management and Sharing Policy. --- ## Quick Answer - Choose your statistical test by first classifying your data type (continuous, ordinal, or categorical) and then counting your groups and whether they are paired or independent. - Check test assumptions including normality and variance homogeneity before running the test, and use the decision tree below to match your study design to the correct procedure. - No single test fits all biological data, and violating assumptions can produce misleading results, so verify your data structure and consult a biostatistician for complex designs. ## Understanding Biological Data Types and Study Designs The foundation of any statistical analysis in biology begins with a clear understanding of your data structure. Biological data come in various forms, and each form requires a different analytical approach. Misclassifying your data type is the most common source of incorrect test selection. ### Continuous Data Continuous data are measurements that can take any value within a range. Examples include body weight in grams, enzyme activity in units per milligram, gene expression levels, and growth rates. These data have meaningful numerical relationships, and the intervals between values are consistent. Continuous data allow for powerful parametric tests when their underlying assumptions are met. ### Categorical Data Categorical data represent membership in a group or category. These data are often counts or frequencies. Examples include the number of plants that survived or died, the presence or absence of a genetic variant, or the classification of animals into treatment groups. Categorical data are analyzed using tests that compare observed frequencies against expected frequencies. ### Ordinal Data Ordinal data fall between continuous and categorical data. These data have a natural order but the intervals between values are not necessarily equal. Examples include disease severity scores, behavioral observation scales, or developmental stages. Ordinal data require special consideration because standard parametric tests assume continuous measurement. ### Paired Versus Independent Observations The relationship between your observations determines whether you use paired or independent tests. Paired observations occur when you measure the same subject before and after treatment, or when you match subjects by important characteristics. Independent observations occur when each subject appears in only one group and there is no natural connection between subjects in different groups. ## The Decision Framework for Test Selection Selecting the right statistical test requires a systematic approach. The decision process follows a logical sequence of questions about your data and study design. ### Step 1: Define Your Research Question Before you select a test, write down your research question in precise terms. Are you comparing groups, examining relationships between variables, or testing whether data fit a predicted distribution? The type of question determines the family of statistical tests you will consider. ### Step 2: Identify Your Variables Determine which variable is your response variable and which are your explanatory variables. The response variable is the outcome you measure. The explanatory variables are the factors you believe influence the response. In a study of fertilizer effects on plant height, plant height is the response variable and fertilizer treatment is the explanatory variable. ### Step 3: Determine the Number of Groups Count the number of groups you are comparing. Some tests handle only two groups, while others handle three or more. Using a two-group test on a three-group design loses information and increases the risk of false positives. ### Step 4: Check the Assumptions Each statistical test has assumptions about the data. The most common assumptions are normality, homogeneity of variance, and independence of observations. You must check these assumptions before running the test. If the assumptions are violated, you may need to use a nonparametric alternative or transform your data. ## At a Glance: Statistical Test Selection Table | Data Type | Number of Groups | Paired or Independent | Recommended Test | | --- | --- | --- | --- | | Continuous | Two groups | Independent | Student's t-test or Welch's t-test | | Continuous | Two groups | Paired | Paired t-test | | Continuous | Three or more | Independent | One-way ANOVA | | Continuous | Three or more | Paired | Repeated measures ANOVA | | Continuous | Two or more | Independent with covariates | ANCOVA | | Categorical | Two groups | Independent | Chi-square test or Fisher's exact test | | Categorical | Two groups | Paired | McNemar's test | | Ordinal or non-normal continuous | Two groups | Independent | Mann-Whitney U test | | Ordinal or non-normal continuous | Two groups | Paired | Wilcoxon signed-rank test | | Ordinal or non-normal continuous | Three or more | Independent | Kruskal-Wallis test | | Ordinal or non-normal continuous | Three or more | Paired | Friedman test | ## Parametric Tests for Continuous Data Parametric tests assume that your data follow a known distribution, typically the normal distribution. These tests are generally more powerful than nonparametric tests when their assumptions are met. ### Student's t-Test for Two Independent Groups The Student's t-test compares the means of two independent groups. This test assumes that the data are normally distributed within each group and that the variances of the two groups are equal. When the variances are not equal, you should use Welch's t-test, which does not assume equal variances. To run a Student's t-test, you need continuous data from two independent groups. The test produces a t-statistic and a p-value. The p-value tells you whether the observed difference between group means is likely to have occurred by chance alone. ### Paired t-Test for Dependent Observations The paired t-test compares the means of two related groups. This test is appropriate when you measure the same subjects under two conditions or when you match subjects by important characteristics. The paired t-test is more powerful than the independent t-test because it removes the variability between subjects. ### One-Way ANOVA for Three or More Independent Groups One-way analysis of variance (ANOVA) compares the means of three or more independent groups. The test determines whether at least one group mean differs from the others. ANOVA produces an F-statistic and a p-value. A significant result tells you that the group means are not all equal, but it does not tell you which groups differ. After a significant ANOVA, you need to run post-hoc tests to identify which groups differ. Common post-hoc tests include Tukey's HSD, Bonferroni correction, and Dunnett's test. The choice of post-hoc test depends on your study design and whether you have a control group. ### Repeated Measures ANOVA for Paired Data with Three or More Groups Repeated measures ANOVA is used when you measure the same subjects under three or more conditions. This test accounts for the correlation between repeated measurements on the same subject. The test is appropriate for longitudinal studies where you measure subjects at multiple time points. ### ANCOVA for Controlling Confounding Variables Analysis of covariance (ANCOVA) extends ANOVA by including one or more continuous covariates. This test allows you to control for the effect of a variable that you cannot manipulate but that influences your response variable. For example, you might control for initial body weight when comparing weight gain across treatment groups. ## Nonparametric Tests for Data That Violate Assumptions Nonparametric tests do not assume a specific distribution for the data. These tests are useful when your data violate the normality assumption or when your data are ordinal. ### Mann-Whitney U Test for Two Independent Groups The Mann-Whitney U test compares the distributions of two independent groups. This test is the nonparametric alternative to the Student's t-test. The test ranks all observations and compares the sum of ranks between the two groups. The test does not require normality but does require that the distributions of the two groups have the same shape. ### Wilcoxon Signed-Rank Test for Two Paired Groups The Wilcoxon signed-rank test compares the distributions of two paired groups. This test is the nonparametric alternative to the paired t-test. The test ranks the absolute differences between paired observations and compares the sum of ranks for positive and negative differences. ### Kruskal-Wallis Test for Three or More Independent Groups The Kruskal-Wallis test compares the distributions of three or more independent groups. This test is the nonparametric alternative to one-way ANOVA. The test ranks all observations and compares the sum of ranks across groups. A significant result indicates that at least one group differs from the others. ### Friedman Test for Three or More Paired Groups The Friedman test compares the distributions of three or more paired groups. This test is the nonparametric alternative to repeated measures ANOVA. The test ranks observations within each subject and compares the sum of ranks across conditions. ## Tests for Categorical Data Categorical data require tests that compare observed frequencies against expected frequencies. These tests are based on the chi-square distribution. ### Chi-Square Test of Independence The chi-square test of independence examines whether two categorical variables are associated. The test compares the observed frequencies in a contingency table against the frequencies expected if the variables were independent. The test requires that the expected frequency in each cell is sufficiently large. ### Fisher's Exact Test for Small Samples Fisher's exact test is used when the sample size is small or when the expected frequencies in a contingency table are too low for the chi-square test. The test calculates the exact probability of observing the data given the marginal totals. This test is particularly useful for 2x2 tables with small counts. ### McNemar's Test for Paired Categorical Data McNemar's test compares paired categorical data. This test is used when you have the same subjects measured under two conditions and the outcome is binary. The test focuses on the discordant pairs, where the outcome differs between the two conditions. ## Correlation and Regression for Relationships Between Variables When your research question involves the relationship between two or more continuous variables, you need correlation or regression analysis. ### Pearson Correlation for Linear Relationships The Pearson correlation coefficient measures the strength and direction of a linear relationship between two continuous variables. The coefficient ranges from -1 to +1. A value of 0 indicates no linear relationship. The test assumes that both variables are normally distributed and that the relationship is linear. ### Spearman Rank Correlation for Nonparametric Relationships The Spearman rank correlation coefficient measures the strength and direction of a monotonic relationship between two variables. This test is the nonparametric alternative to the Pearson correlation. The test ranks the data and calculates the correlation on the ranks. The test does not require normality and can detect monotonic relationships that are not linear. ### Simple Linear Regression for Predicting Outcomes Simple linear regression models the relationship between one response variable and one explanatory variable. The regression equation describes how the response variable changes with the explanatory variable. The test produces an R-squared value that indicates the proportion of variance in the response variable explained by the explanatory variable. ### Multiple Regression for Multiple Predictors Multiple regression extends simple linear regression to include two or more explanatory variables. This test allows you to examine the effect of each predictor while controlling for the other predictors. The test produces coefficients for each predictor and an overall model fit statistic. ## Practical Workflow for Test Selection The following workflow guides you through the process of selecting and running the appropriate statistical test for your biological data. ### Step 1: Organize Your Data Organize your data in a spreadsheet with each row representing a subject or observation and each column representing a variable. Ensure that your data are clean and that missing values are handled appropriately. Record your data in a consistent format that your statistical software can read. ### Step 2: Classify Your Variables Classify each variable as continuous, ordinal, or categorical. Determine which variable is your response variable and which are your explanatory variables. This classification determines the family of tests you will consider. ### Step 3: Determine the Study Design Determine whether your observations are paired or independent. Identify the number of groups you are comparing. This information narrows your test options to a specific set of tests. ### Step 4: Check the Assumptions Run the assumption checks for the candidate tests. For parametric tests, check normality and homogeneity of variance. Use the Shapiro-Wilk test for normality and Levene's test for homogeneity of variance. If the assumptions are violated, consider nonparametric alternatives or data transformations. ### Step 5: Run the Test and Interpret the Results Run the selected test and interpret the results in the context of your research question. Report the test statistic, degrees of freedom, and p-value. Interpret the p-value in the context of your significance level, typically 0.05. ### Step 6: Document Your Analysis Document your analysis steps, including the tests you considered, the assumption checks you ran, and the final test you selected. This documentation supports the reproducibility of your analysis and helps reviewers understand your decisions. ## Common Failure Patterns in Test Selection Several common errors lead to incorrect test selection and misleading results. ### Using Parametric Tests on Non-Normal Data Parametric tests assume normality. When the data are not normally distributed, the test results may be unreliable. The t-test and ANOVA are robust to moderate violations of normality when the sample size is large, but they can produce misleading results with small samples or severe violations. ### Ignoring the Paired Structure of Data When observations are paired, you must use a paired test. Using an independent test on paired data ignores the correlation between observations and reduces the power of the test. This error can lead to false negative results. ### Running Multiple Tests Without Correction When you compare three or more groups, you should use ANOVA instead of running multiple t-tests. Running multiple t-tests increases the risk of false positives. If you must run multiple comparisons, apply a correction such as the Bonferroni correction. ### Using the Chi-Square Test with Small Expected Frequencies The chi-square test requires that the expected frequencies in each cell be sufficiently large. When the expected frequencies are too small, the test may produce inaccurate results. In this case, use Fisher's exact test instead. ### Ignoring the Assumption of Independence Many statistical tests assume that observations are independent. When observations are correlated, such as when you have multiple measurements from the same subject, you must account for the correlation. Ignoring the correlation can lead to inflated test statistics and false positives. ## Reproducibility and Reporting of Statistical Analyses Transparent reporting of your statistical analysis is essential for the credibility of your research. The EQUATOR Network provides reporting guidelines for various study designs, and you should consult these guidelines when preparing your manuscript. ### Reporting the Test and Assumptions Your research report should describe the statistical tests you used and the assumptions you checked. Include the test statistic, degrees of freedom, and p-value for each test. Describe how you handled missing data and any data transformations you applied. ### Reporting Effect Sizes A p-value tells you whether a result is statistically significant, but it does not tell you the magnitude of the effect. Report effect sizes such as Cohen's d for t-tests or eta-squared for ANOVA. Effect sizes help readers understand the practical importance of your findings. ### Sharing Data and Code Sharing your data and analysis code supports the reproducibility of your research. The National Institutes of Health Data Management and Sharing Policy describes expectations for sharing data and supporting documentation. Plan for data sharing when you design your study and include a data management plan in your grant application. ### Registering Your Study For clinical and some biological studies, registering your study before you begin can improve transparency and reduce the risk of selective reporting. The EQUATOR Network provides resources for reporting guidelines and study registration. ## Software Options for Statistical Analysis Several software packages can perform the statistical tests described in this article. The choice of software depends on your preferences, your institution's resources, and the complexity of your analysis. ### R for Statistical Computing R is a free and open-source programming language for statistical computing. R provides a wide range of statistical tests and visualization tools. The software is highly extensible, and you can find packages for almost any statistical analysis. R requires some programming knowledge, but the learning curve is manageable for most researchers. ### SPSS for Point-and-Click Analysis SPSS provides a point-and-click interface for statistical analysis. The software is widely used in the social and biological sciences. SPSS provides a range of tests and produces output in a format that is easy to interpret. The software is commercial and requires a license. ### Python for Statistical Analysis Python is a general-purpose programming language that provides statistical analysis through libraries such as SciPy and statsmodels. Python is useful when you need to integrate statistical analysis with other data processing tasks. The software is free and open-source. ## Practical Implementation Steps for Your Research ### Step 1: Write Your Research Question Write your research question in a clear and precise format. Identify the response variable and the explanatory variables. Determine whether you are comparing groups, examining relationships, or testing a distribution. ### Step 2: Design Your Study Design your study to answer your research question. Determine the sample size you need to detect the effect you are interested in. Consider whether your observations will be paired or independent and how many groups you will compare. ### Step 3: Collect and Organize Your Data Collect your data according to your study design. Organize your data in a spreadsheet or database with clear variable names and consistent coding. Record your data in a format that your statistical software can read. ### Step 4: Explore Your Data Before running any statistical tests, explore your data visually and numerically. Create histograms and boxplots to examine the distribution of your variables. Calculate summary statistics such as the mean, median, and standard deviation. ### Step 5: Select and Run the Test Use the decision table in this article to select the appropriate test for your data. Run the assumption checks and then run the test. Record the test statistic, degrees of freedom, and p-value. ### Step 6: Interpret the Results Interpret the results in the context of your research question. Consider the magnitude of the effect and the practical significance of the findings. Report the results in your research paper or report. ### Step 7: Document and Share Your Analysis Document your analysis steps and share your data and code when possible. This documentation supports the reproducibility of your research and helps other researchers understand your methods. ## Limitations and Considerations ### Sample Size Considerations The sample size affects the power of your statistical test. A small sample size may not detect a real effect, while a large sample size may detect effects that are not biologically meaningful. Consider the effect size you are interested in and the variability of your data when determining the sample size. ### Multiple Testing When you run multiple statistical tests, the risk of false positives increases. Use a correction method such as the Bonferroni correction or the false discovery rate to control the risk of false positives. Consider whether you need to correct for multiple testing in your study design. ### Data Transformations When your data violate the assumptions of a parametric test, you may be able to transform the data to meet the assumptions. Common transformations include the log transformation, the square root transformation, and the arcsine transformation. Transformations can make the data more normal and stabilize the variance. ### Missing Data Missing data can affect the validity of your statistical analysis. Consider how you will handle missing data before you begin your analysis. Options include complete-case analysis, imputation, and the use of statistical methods that handle missing data. ## When to Consult a Biostatistician Some study designs require the expertise of a biostatistician. You should consider consulting a biostatistician when your study design is complex, when you have multiple variables, or when you are unsure about the appropriate test. ### Complex Study Designs Study designs with multiple factors, repeated measurements, or nested structures require specialized statistical methods. A biostatistician can help you select the appropriate model and interpret the results. ### High-Stakes Decisions When your research results will inform important decisions, such as regulatory submissions or clinical practice, you should consult a biostatistician. The biostatistician can help you ensure that your analysis is valid and that your conclusions are supported by the data. ### Unclear Data Structure When you are unsure about the structure of your data or the assumptions of your test, consult a biostatistician. They can help you clarify your data structure and select the appropriate test. ## Building a Test Selection Audit Trail for Your Biological Data Choosing the correct statistical test is only half of the analytical challenge. The other half is documenting why you made that choice so that your analysis can withstand scrutiny from reviewers, regulators, and other researchers who may want to reproduce your work. A test selection audit trail is a structured record that captures every decision point in your analytical workflow, from raw data structure through assumption checks to the final test selection. This record transforms a potentially opaque analytical process into a transparent, verifiable chain of reasoning. ### Why You Need an Audit Trail Biological research increasingly demands transparency in data analysis. The National Institutes of Health Data Management and Sharing Policy describes expectations for sharing data and supporting documentation, and this policy applies to the analytical decisions you make as well as the raw data themselves. When you share your data, reviewers and other researchers need to understand why you selected a particular test. Without an audit trail, they cannot verify that your test choice was appropriate for your data structure. The Committee on Publication Ethics Core Practices emphasize the importance of data integrity and transparency in research. An audit trail directly supports these practices by documenting the analytical decisions that lead to your conclusions. If a question arises about your analysis after publication, the audit trail provides the evidence needed to address that question. An audit trail also protects you from common analytical errors. When you write down each decision, you are forced to think through your reasoning. This process often reveals mistakes that you might otherwise overlook, such as using a paired test on independent data or applying a parametric test to data that violate the normality assumption. ### What to Record in Your Audit Trail Your audit trail should capture every decision point in your test selection process. The following records form the core of a complete audit trail. #### Data Structure Record The first entry in your audit trail is a description of your data structure. Record the type of each variable in your dataset. For each variable, note whether it is continuous, ordinal, or categorical. Identify your response variable and your explanatory variables. Record the number of groups you are comparing and whether the observations are paired or independent. This record should be created before you run any statistical tests. It serves as the foundation for all subsequent decisions. If you need to revise your data structure later, record the revision and the reason for it. #### Assumption Check Record For each candidate test, record the assumption checks you performed. Include the name of each test you used to check assumptions, the test statistic, and the p-value. For example, if you used the Shapiro-Wilk test to check normality, record the W statistic and the p-value for each group. If you used Levene's test for homogeneity of variance, record the F statistic and the p-value. Record the date you ran each assumption check and the software you used. This information allows someone else to reproduce your assumption checks and verify your conclusions. #### Test Selection Record Record the test you selected and the reason for your selection. Note the alternative tests you considered and why you rejected them. For example, if you chose the Mann-Whitney U test over the Student's t-test because your data violated the normality assumption, record the normality test result that led to this decision. This record should also include the test statistic, degrees of freedom, and p-value from the final test. Record the effect size and the confidence interval if you calculated them. #### Data Transformation Record If you transformed your data, record the transformation you applied and the reason for the transformation. Record the original data values and the transformed values. This allows someone else to verify that the transformation was appropriate and that the analysis was performed on the correct data. #### Missing Data Record Record how you handled missing data. Note the number of missing values in each variable and the method you used to handle them. If you used complete-case analysis, record the number of cases you excluded. If you used imputation, record the imputation method and the software you used. ### Building the Audit Trail in Practice The audit trail should be built as you work, not after you have completed your analysis. The following steps describe how to build the audit trail alongside your analysis. #### Step 1: Create the Audit File Create a new document for your audit trail. This can be a text file, a spreadsheet, or a section in your laboratory notebook. The format does not matter as long as you can record all the necessary information. Include the date and a description of the dataset you are analyzing. #### Step 2: Record Your Data Structure Before you run any tests, record your data structure. List each variable and its type. Identify the response variable and the explanatory variables. Record the number of groups and whether the observations are paired or independent. This record should be complete before you proceed to assumption checks. #### Step 3: Run and Record Assumption Checks Run the assumption checks for your candidate tests. Record the test name, the test statistic, the p-value, and the date. Record the software and version you used. If you used R, record the package and function. If you used SPSS, record the procedure name. #### Step 4: Record Your Test Selection After you have completed the assumption checks, record the test you selected. Write a brief justification for your selection. Include the alternative tests you considered and why you rejected them. Record the test statistic, degrees of freedom, and p-value from the final test. #### Step 5: Record Any Transformations If you transformed your data, record the transformation and the reason for it. Record the original values and the transformed values. This record is important because transformations can affect the interpretation of your results. #### Step 6: Record Missing Data Handling Record how you handled missing data. Note the number of missing values and the method you used. If you used imputation, record the imputation method and the software you used. #### Step 7: Review and Finalize Review your audit trail to ensure that it is complete and accurate. Check that you have recorded all the necessary information and that the records are consistent with your analysis. Finalize the audit trail and store it with your data and analysis code. ## Common Failure Patterns in Audit Trail Creation Several common errors can undermine the usefulness of an audit trail. ### Creating the Audit Trail After the Analysis The most common error is creating the audit trail after the analysis is complete. This approach is problematic because you may not remember all the decisions you made and the reasons for them. You may also be tempted to revise your decisions to match the results you obtained. Create the audit trail as you go, not after the fact. ### Omitting Assumption Check Results Some researchers record the final test but omit the assumption checks. This omission makes it impossible to verify that the test selection was appropriate. Always record the assumption checks and their results. ### Failing to Record Alternative Tests When you select a test, you should record the alternatives you considered and why you rejected them. This information is valuable because it shows that you considered the full range of options and made an informed decision. ### Not Recording Software Versions Statistical software changes over time, and different versions may produce different results. Record the software name and version you used. This allows someone else to reproduce your analysis with the same software version. ### Ignoring the Audit Trail for Exploratory Analysis Exploratory analysis is often informal and iterative. However, if you use the results of exploratory analysis to make decisions about your formal analysis, you should record those decisions. For example, if you use a histogram to decide whether to transform your data, record that decision and the reason for it. ## Using the Audit Trail for Reproducibility The audit trail is a key component of a reproducible analysis. When you share your data and analysis code, you should also share the audit trail. This allows other researchers to understand your analytical decisions and to verify that your analysis was appropriate. The National Institutes of Health Data Management and Sharing Policy describes expectations for sharing data and supporting documentation. The audit trail is part of the supporting documentation. When you prepare your data management plan, include the audit trail as a component of your documentation. The EQUATOR Network provides reporting guidelines for various study designs. These guidelines often require you to describe your statistical methods in detail. The audit trail provides the information you need to complete these descriptions accurately. ### Sharing the Audit Trail When you share your audit trail, use a format that is accessible to other researchers. A text file or a spreadsheet is usually sufficient. Include the audit trail with your data and analysis code in a repository or as supplementary material. ### Using the Audit Trail in Peer Review The audit trail can be used during peer review to verify your analytical decisions. Reviewers can check that you selected the appropriate test and that you checked the assumptions. This verification strengthens the credibility of your research. ## The Audit Trail as a Teaching Tool The audit trail is also a valuable teaching tool. When you teach students how to select statistical tests, you can use the audit trail to show them the decision process. The audit trail makes the decision process explicit and concrete, which helps students understand the reasoning behind test selection. The audit trail can also be used to teach the importance of reproducibility. By showing students how to record their analytical decisions, you teach them a skill that will serve them throughout their research careers. ## Integrating the Audit Trail with Your Research Workflow The audit trail should be integrated into your research workflow from the beginning. When you design your study, plan how you will record your analytical decisions. When you collect your data, record the data structure. When you analyze your data, record the assumption checks and the test selection. When you share your data, share the audit trail. The audit trail is not a separate activity. It is a part of the analysis itself. By building the audit trail as you go, you create a complete record of your analytical decisions that supports the reproducibility and credibility of your research. ## The Audit Trail and the Decision Framework The audit trail complements the decision framework described in the main article. The decision framework tells you which test to select. The audit trail records why you selected that test. Together, they provide a complete picture of your analytical process. The decision framework and the audit trail are both essential for a rigorous analysis. The decision framework ensures that you select the appropriate test. The audit trail ensures that you can justify your selection to others. ## The Audit Trail and the Reporting Guidelines The EQUATOR Network provides reporting guidelines for various study designs. These guidelines often require you to describe your statistical methods in detail. The audit trail provides the information you need to meet these requirements. For example, the guidelines may require you to describe how you checked the assumptions of your test. The audit trail records the assumption checks you performed and the results. This information allows you to describe your methods accurately and completely. ## The Audit Trail and the Data Management Plan The National Institutes of Health Data Management and Sharing Policy requires you to include a data management plan in your grant application. The plan should describe how you will manage and share your data. The audit trail is part of the data management plan. It describes how you will document your analytical decisions. When you write your data management plan, include a section on the audit trail. Describe the format you will use and the information you will record. This section shows that you have planned for the reproducibility of your analysis. ## The Audit Trail and the Publication Process The audit trail can be used during the publication process. When you submit your manuscript, you can include the audit trail as supplementary material. This allows reviewers to verify your analytical decisions. The audit trail can also be used to respond to reviewer comments about your statistical methods. The Committee on Publication Ethics Core Practices emphasize the importance of data integrity. The audit trail supports this by providing a record of your analytical decisions. When you publish your research, you can share the audit trail to demonstrate the integrity of your analysis. ## The Audit Trail and the Research Community The audit trail is a valuable contribution to the research community. When you share your audit trail, you help other researchers understand your analysis. This understanding is essential for the reproducibility of your research. The audit trail also helps other researchers learn from your experience. By seeing how you made your analytical decisions, they can learn how to make their own decisions. This learning is important for the development of the research community. ## The Audit Trail and the Future of Research The audit trail is part of a broader movement toward transparent and reproducible research. As the research community continues to emphasize reproducibility, the audit trail will become an increasingly important part of the research process. By building the audit trail into your workflow, you are preparing for the future of research. The audit trail is a simple but powerful tool. It is a record of your analytical decisions that you can use to verify your analysis and to share with others. By building the audit trail, you are making your research more transparent and more credible. ## Frequently Asked Questions ### What is the first step in choosing a statistical test? The first step is to classify your data type as continuous, ordinal, or categorical. Then identify your response variable and explanatory variables. This classification determines which family of tests you will consider. ### When should I use a nonparametric test? Use a nonparametric test when your data violate the assumptions of a parametric test, such as normality or homogeneity of variance. Nonparametric tests are also appropriate for ordinal data. ### What is the difference between paired and independent tests? Paired tests are used when the observations are related, such as when you measure the same subjects before and after treatment. Independent tests are used when the observations are unrelated and each subject appears in only one group. ### How do I choose between the chi-square test and Fisher's exact test? Use the chi-square test when the expected frequencies in each cell of the contingency table are sufficiently large. Use Fisher's exact test when the expected frequencies are too small or when the sample size is small. ### What is a post-hoc test and when do I need one? A post-hoc test is used after a significant ANOVA to identify which groups differ from each other. You need a post-hoc test when you have three or more groups and your ANOVA is significant. ### How do I check the normality assumption? You can check the normality assumption using the Shapiro-Wilk test or by examining a histogram or Q-Q plot of the data. The Shapiro-Wilk test provides a p-value that indicates whether the data are normally distributed. ### What should I report when I describe my statistical analysis? Report the test statistic, degrees of freedom, and p-value for each test. Also report the effect size and describe how you handled missing data and any data transformations. ### When should I consult a biostatistician? Consult a biostatistician when your study design is complex, when you have multiple variables, or when you are unsure about the appropriate test. A biostatistician can help you design your study and analyze your data correctly. ## 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 - [Selecting Persistent Identifiers for Research Data: A Decision Framework](/knowledge/bioinformatics/selecting-persistent-identifiers-for-research-data-a-decision-framework) - [Metagenomics vs Metabarcoding: Choosing the Right Approach for Your Study](/knowledge/bioinformatics/metagenomics-vs-metabarcoding-choosing-the-right-approach-for-your-study) - [Genomic Data Types: A Primer for Bioinformatics Beginners](/knowledge/bioinformatics/genomic-data-types-a-primer-for-bioinformatics-beginners) - [Metabolomics Data Analysis Workflow: From Raw Data to Biological Insight](/knowledge/bioinformatics/metabolomics-data-analysis-workflow-from-raw-data-to-biological-insight) - [Genomic Diagnostics: Choosing the Right Test for Clinical and Veterinary Applications](/knowledge/bioinformatics/genomic-diagnostics-choosing-the-right-test-for-clinical-and-veterinary-applications) ## Related Clinical & Scientific Guides * [A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data](/knowledge/bioinformatics/a-practical-guide-to-detecting-antimicrobial-resistance-genes-in-shotgun-metagenomic-data) * [Computational Immunology: Modeling the Immune System](/knowledge/bioinformatics/computational-immunology-modeling-the-immune-system) * [How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices](/knowledge/bioinformatics/how-to-set-hard-filters-for-germline-variant-calling-a-practical-guide-to-gatk-best-practices) ## 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. - [Prothrombin Complex Concentrate vs Frozen Plasma for Coagulopathic Bleeding in Cardiac Surgery: The FARES-II Multicenter Randomized Clinical Trial.](https://pubmed.ncbi.nlm.nih.gov/40156829). JAMA, 2025. - [Six persistent research misconceptions.](https://pubmed.ncbi.nlm.nih.gov/24452418). Journal of general internal medicine, 2014. - [From genome-wide associations to candidate causal variants by statistical fine-mapping.](https://pubmed.ncbi.nlm.nih.gov/29844615). Nature reviews. Genetics, 2018. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.