Statistical Methods in the Methods

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

Quick Answer

• Read the statistical methods section to verify that each test matches the research question, data type, and study design before trusting the results.
• Check for normality testing, multiple comparison corrections, and sample size justification as the three most common decision points.
• A methods section cannot prove a study is correct, it only shows what the authors claim they did, so compare it against reporting guidelines and raw data when available.

Why Statistical Methods Matter in Animal Research

Animal farming decisions increasingly depend on published research. Whether you are evaluating a new feed additive, a vaccination protocol, or a housing system change, the statistical methods section determines whether the conclusions deserve your attention. A study with poor statistical choices can produce misleading results that look convincing on the surface.

The statistical methods section is not a formality. It is the part of a paper where authors explain how they converted raw observations into conclusions. For farmers and animal owners, this section answers a practical question: can I expect the same result if I apply this treatment to my herd or flock?

Researchers in animal science face the same reporting pressures as other fields. The Committee on Publication Ethics Core Practices emphasize that research should be reported honestly and transparently, with proper handling of data and authorship. When statistical methods are described poorly, readers cannot tell whether the authors followed those standards.

The Role of Reporting Guidelines

Reporting guidelines exist to help authors describe their methods completely and accurately. The EQUATOR Network maintains a collection of these guidelines for different study types. For animal research, the relevant guideline depends on the study design, such as randomized trials, observational studies, or diagnostic accuracy studies.

When you read a statistical methods section, check whether the authors mention a reporting guideline. This mention signals that the authors intended to follow a recognized standard for transparency. The EQUATOR Network provides a searchable library where you can find the specific guideline that applies to the paper you are reading.

A methods section that follows a reporting guideline will typically include information about sample size calculation, randomization procedures, blinding, and statistical analysis plans. These elements allow you to assess whether the study was designed to answer its research question without bias.

Core Components of a Statistical Methods Section

Study Design and Unit of Analysis

The first thing to identify is the experimental unit. In animal research, the experimental unit is the smallest unit that receives a treatment independently. This could be an individual animal, a pen, a barn, or a pasture. A common error occurs when researchers treat individual animals as independent when they were housed together in groups.

The unit of analysis must match the experimental unit. If pigs are housed in pens and each pen receives a different feed treatment, then the pen is the experimental unit. Analyzing individual pig weights as if each pig were independent inflates the sample size and can create false positive results.

Look for a clear statement about how animals were assigned to treatment groups. Random assignment helps ensure that groups are comparable at the start of the study. The methods should describe the randomization method, such as random number generation or block randomization.

Sample Size Justification

A proper methods section explains how the sample size was determined. This explanation usually includes the expected effect size, the acceptable error rate, and the statistical power the study aimed to achieve. Power is the probability that the study will detect a true effect if one exists.

Without a sample size justification, you cannot tell whether the study had enough animals to detect a meaningful difference. Underpowered studies may miss real effects, while overpowered studies may detect tiny differences that have no practical value for your operation.

For animal research, sample size calculations should account for expected mortality or culling. If the authors did not plan for losses, the final sample size may be smaller than intended, reducing the study's ability to detect differences.

Normality Testing and Data Distribution

Many common statistical tests assume that data follow a normal distribution. The methods section should state whether the authors tested this assumption and how. Common approaches include the Shapiro-Wilk test, the Kolmogorov-Smirnov test, or visual inspection of histograms and Q-Q plots.

When data are not normally distributed, authors should use nonparametric tests or transform the data. Nonparametric tests, such as the Mann-Whitney U test or Kruskal-Wallis test, do not require normality assumptions. Alternatively, authors may apply logarithmic or square root transformations to make the data more normal.

For farmers, the practical implication is straightforward. If the methods do not mention normality testing, the authors may have used an inappropriate test. This does not automatically invalidate the results, but it warrants caution when interpreting the findings.

Choice of Statistical Tests

The methods section should name the specific tests used for each comparison. Common tests in animal research include:

• t-tests for comparing two group means
• Analysis of variance (ANOVA) for comparing three or more group means
• Chi-square tests for categorical outcomes
• Regression analysis for relationships between variables
• Survival analysis for time-to-event outcomes such as mortality

Each test has assumptions that must be met. The methods should explain why a particular test was chosen and how the data satisfied the test's requirements. If the authors used multiple tests, the methods should describe how they handled the risk of false positives from multiple comparisons.

Multiple Comparisons and Adjustments

When researchers test many hypotheses at once, the chance of finding a significant result by chance increases. The methods section should describe how the authors controlled for this problem. Common adjustment methods include the Bonferroni correction, the Tukey honest significant difference test, and the false discovery rate approach.

For example, a study comparing four different feed formulations might test six pairwise comparisons. Without adjustment, the chance of at least one false positive is much higher than the nominal 5 percent error rate. The methods should state which adjustment was used and why.

If the methods section does not mention multiple comparison adjustments, the reported p-values may be misleading. This is especially important in studies with many outcome measures, such as growth performance, carcass quality, and blood parameters.

At a Glance

How to Read a Statistical Methods Section

Step 1: Identify the Research Question

Before evaluating the statistics, determine what the study aimed to measure. The research question should be stated clearly in the introduction and reflected in the methods. Write down the primary outcome and any secondary outcomes. This gives you a framework for evaluating whether the statistical tests match the question.

Step 2: Locate the Experimental Unit

Find the sentence that describes how animals were allocated to treatments. If the paper does not state this clearly, look for clues in the housing description. Animals in the same pen or same barn are likely not independent observations. The statistical analysis should account for this grouping.

Step 3: Check the Sample Size Logic

Look for a statement about how many animals were needed and why. The justification should include the expected difference between groups and the variability the authors anticipated. If the sample size seems arbitrary, the study may not have enough power to support its conclusions.

Step 4: Evaluate the Test Selection

For each outcome, identify the test used and ask whether it matches the data type. Continuous outcomes like weight gain typically use t-tests or ANOVA. Categorical outcomes like mortality rates use chi-square tests or logistic regression. If the test does not match the data type, the results may be unreliable.

Step 5: Look for Adjustment Methods

Count how many comparisons the authors made. If they tested many outcomes or many group pairs, check for a multiple comparison adjustment. The absence of adjustment does not always invalidate the study, but it means the p-values should be interpreted with more caution.

Step 6: Verify Software and Reproducibility

The methods should name the statistical software and version used for analysis. This detail matters because different software versions can produce slightly different results. The methods should also describe any data cleaning or exclusion criteria so you know exactly which animals were included in the analysis.

Common Statistical Tests in Animal Research

t-Tests

The t-test compares the means of two groups. It is appropriate when the outcome is continuous, the data are approximately normal, and the groups are independent. A paired t-test is used when the same animals are measured before and after a treatment.

In animal research, t-tests often compare a treatment group against a control group. The methods should state whether the test was one-tailed or two-tailed. A one-tailed test assumes the treatment can only work in one direction, which is rarely justified in animal research.

Analysis of Variance

ANOVA extends the t-test to three or more groups. It tests whether any group mean differs from the others. A significant ANOVA result does not tell you which groups differ, so follow-up tests are needed. The methods should describe these follow-up comparisons.

ANOVA assumes homogeneity of variance, meaning the variability is similar across groups. The methods should mention whether this assumption was tested, often with Levene's test or Bartlett's test. If variances are unequal, the authors should use a Welch ANOVA or a nonparametric alternative.

Chi-Square Tests

Chi-square tests analyze categorical data, such as the number of animals that died or developed a disease in each treatment group. The test compares observed counts against expected counts under the null hypothesis.

The methods should state whether the expected cell counts met the minimum requirements for the chi-square test to be valid. When expected counts are low, Fisher's exact test is more appropriate. If the methods do not address this, the chi-square results may be unreliable.

Regression Analysis

Regression models examine relationships between variables. Linear regression predicts a continuous outcome from one or more predictor variables. Logistic regression predicts a binary outcome, such as whether an animal survives or develops a disease.

In animal research, regression models often include covariates such as initial body weight, age, or sex. The methods should list which covariates were included and how they were selected. The methods should also describe how the authors checked model assumptions, such as linearity and independence of errors.

Mixed Models

Mixed models account for both fixed effects and random effects. Fixed effects are the treatments you are testing. Random effects account for variation from grouping factors, such as pen, barn, or farm. Mixed models are increasingly common in animal research because they handle clustered data correctly.

The methods should describe the random effects structure and the covariance structure used in the model. If the authors used a mixed model, they should explain why this approach was necessary and how they selected the final model.

Practical Assessment Steps for Readers

Create a Methods Checklist

Before reading the results section, prepare a checklist based on the elements described above. For each element, note whether the paper provides adequate information. This checklist helps you identify gaps in the methods before the results influence your judgment.

Compare Methods Against Results

After reading the methods, check whether the results section reports the same analyses. The number of animals analyzed should match the sample size described in the methods. The tests named in the methods should appear in the results. Discrepancies between methods and results may indicate selective reporting.

Look for Supplementary Materials

Many journals publish supplementary files with additional methodological detail. These files may include the full statistical analysis plan, raw data, or analysis code. The National Institutes of Health Data Management and Sharing Policy emphasizes the importance of data sharing for research transparency. If the paper mentions data availability, check whether the data are accessible.

Consider the Funding Source

The methods section or the acknowledgments may disclose funding sources. Funding from feed companies or pharmaceutical manufacturers does not automatically invalidate a study, but it provides context for interpreting the results. The Committee on Publication Ethics Core Practices require authors to declare conflicts of interest, so check whether a conflict statement is present.

Records and Measurements to Track

When you evaluate a statistical methods section, keep a record of your assessment. This record helps you compare studies over time and identify patterns in the literature you rely on for management decisions.

Common Failure Patterns in Statistical Methods

Vague Test Descriptions

Some papers state that "appropriate statistical tests were used" without naming the tests. This pattern prevents readers from evaluating the analysis. If the authors do not specify their methods, you cannot verify that the tests matched the data type and study design.

Ignoring Nonindependence

Animals housed together share environmental conditions that make their outcomes correlated. When authors treat each animal as independent despite group housing, the analysis overstates the precision of the results. This pattern is common in pen-based feeding trials and group-housed disease studies.

Selective Reporting of Outcomes

Authors may describe methods for several outcomes but only report results for the ones that reached statistical significance. This pattern, sometimes called outcome reporting bias, makes the literature appear more consistent than it actually is. Comparing the methods section against the results section helps you detect this problem.

Inadequate Handling of Missing Data

Animal studies frequently lose animals to death, culling, or equipment failure. The methods should describe how missing data were handled. Common approaches include complete case analysis, last observation carried forward, or multiple imputation. If the methods do not address missing data, the results may be biased.

Overreliance on P-Values

Some methods sections focus exclusively on p-values without reporting effect sizes or confidence intervals. A p-value tells you whether a difference is likely due to chance, but it does not tell you the size of the difference. For farming decisions, the effect size matters more than statistical significance. A statistically significant difference of 1 gram per day in weight gain may have no economic value.

Welfare and Safety Context

Statistical methods have direct implications for animal welfare and safety. Studies with inadequate sample sizes may fail to detect adverse effects of a treatment. Studies with inappropriate statistical tests may overstate the benefits of a product while missing its risks.

When you evaluate a study that will inform your animal management decisions, consider the welfare consequences of being wrong. If a study claims a new vaccine reduces mortality but used an underpowered design, the vaccine may not perform as expected in your herd. If a study claims a feed additive improves growth but ignored pen effects, the additive may not work under your housing conditions.

The National Library of Medicine Research Methods Resources provide access to authoritative references on research design and analysis. These resources can help you understand the statistical concepts that appear in animal research papers.

Limitations of Statistical Methods Reading

Reading the statistical methods section has inherent limitations. The methods describe what the authors claim they did, not necessarily what they actually did. Without access to the raw data and analysis code, you cannot fully verify the analysis.

The National Institutes of Health Data Management and Sharing Policy encourages researchers to share data and code, but not all studies comply. When data are not available, you must rely on the methods description and your assessment of the authors' credibility.

Statistical methods reading also requires some statistical literacy. If you are unfamiliar with terms like p-values, confidence intervals, or mixed models, consider consulting educational resources or collaborating with a statistician. The National Library of Medicine bookshelf provides introductory materials on biostatistics and research methods.

Professional Escalation Criteria

Some situations warrant consulting a statistician or research methodologist. Consider professional input when:

• The methods section uses unfamiliar statistical techniques that you cannot evaluate
• The study results will inform a major management decision with significant financial or welfare consequences
• Multiple studies on the same topic report conflicting results and you need to determine which are more reliable
• You are designing your own research or on-farm trial and need guidance on sample size and analysis
• A regulatory or certification body requires evidence that a product or practice meets specific standards

A statistician can help you interpret complex methods and assess whether the analysis supports the conclusions. This consultation is especially valuable when you are considering adopting a new practice based on published research.

Frequently Asked Questions

What is the most important part of a statistical methods section?

The most important part is the description of the experimental unit and how animals were assigned to treatments. This determines whether the entire analysis is valid. If the experimental unit is wrong, every p-value and confidence interval in the paper is suspect.

How do I know if a sample size is adequate?

Look for a power calculation that states the expected effect size, the variability, and the desired power. The calculation should also account for expected losses. If the paper does not provide this information, you cannot determine whether the sample size was adequate.

What does normality testing mean and why does it matter?

Normality testing checks whether data follow a normal bell-shaped distribution. Many common tests like t-tests and ANOVA assume normality. If the data are not normal and the authors did not use alternative tests or transformations, the results may be unreliable.

Why do multiple comparison adjustments matter?

When researchers test many hypotheses, the chance of finding a significant result by chance increases. Adjustments like the Bonferroni correction reduce this risk. Without adjustments, some significant results may be false positives.

Can I trust a paper that does not mention reporting guidelines?

A missing reporting guideline mention does not automatically invalidate a study. However, it reduces your ability to verify that the methods were complete. You should look more carefully for missing methodological details.

What should I do if the methods and results do not match?

Compare the number of animals analyzed and the tests reported in both sections. If they differ, contact the corresponding author for clarification. If the discrepancy is substantial, consider whether the study should influence your decisions.

How can I learn more about statistical methods?

The National Library of Medicine Research Methods Resources provide access to educational materials on research design and statistics. You can also consult textbooks on biostatistics or take online courses in research methods.

When should I consult a statistician?

Consult a statistician when you need to evaluate complex methods, design your own study, or make a major decision based on conflicting research. Professional statistical input is a reasonable investment when the stakes are high.

Related Articles

• Writing the Methods Section for Qualitative Research Papers
• Writing the Methods Section of a Scientific Paper: A Reproducibility-Focused Guide
• Statistical Analysis Methods
• Trend Analysis in Research: Methods, Applications, and Pitfalls
• Network Meta-Analysis: Methods and Applications for Comparative Effectiveness

References and Further Reading

• Research Methods Resources. National Library of Medicine.
• EQUATOR Network. EQUATOR Network.
• Core Practices. Committee on Publication Ethics.
• NIH Grants and Funding. National Institutes of Health.
• ORCID for Researchers. ORCID.
• Data Management and Sharing Policy. National Institutes of Health.
• NCBI Literature Resources. National Center for Biotechnology Information.
• PubMed. National Library of Medicine.

This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.