Common statistical mistakes in student papers: what reviewers find and how to fix them

Statistical errors are the single most frequent reason peer reviewers return student research papers for major revision. This post is written for high school students who have conducted quantitative research and want to understand what those errors are, why they occur, and how to correct them before submission. After reading, you will be able to audit your own methods section and results section for the most consequential errors. If your paper is ready for peer review after that audit, the Princeton Journal of Pre-Collegiate Research publishes original student research across all academic disciplines.

What are the most common statistical mistakes in student papers?

The most common statistical mistakes in student research papers are: using the wrong statistical test for the data type, failing to check test assumptions before running an analysis, misinterpreting p-values as proof of an effect, reporting results without effect sizes, and drawing causal conclusions from correlational data. Each of these errors is detectable in peer review and each one is correctable before submission.

Quantitative research requires two things that are distinct from each other: collecting data correctly and analysing it correctly. Many student papers succeed at the first and fail at the second. The errors listed above are not signs of weak intelligence. They are signs of incomplete statistical training, which is a reasonable limitation for a high school researcher. The goal of this post is to close that gap.

Understanding these errors matters for a practical reason. A paper with a strong research question and well-collected data can be rejected at the peer review stage because the statistical analysis does not support the conclusions drawn. Fixing the analysis does not require re-running the study. In most cases, it requires re-running the tests with the correct parameters and rewriting the results section to reflect what the data actually show.

The five errors below account for the large majority of statistical problems identified in student submissions. Each one is explained with a concrete example and a specific fix.

1. Using the wrong statistical test

The choice of statistical test depends on the type of data collected, the number of groups being compared, and whether the data meet certain distributional assumptions. A student comparing test scores between two groups and using a chi-square test when a t-test is appropriate, for example, will produce results that are not interpretable. Chi-square tests are designed for categorical data. T-tests are designed for continuous data with two groups. Using the wrong test produces a number, but that number does not answer the research question.

Before selecting a test, identify: (a) whether the outcome variable is categorical or continuous, (b) how many groups are being compared, and (c) whether the data are paired or independent. These three questions narrow the choice to a small set of appropriate options.

2. Failing to check test assumptions

Every parametric statistical test rests on assumptions about the data. A t-test assumes that the data are approximately normally distributed. A Pearson correlation assumes a linear relationship between variables. An ANOVA assumes homogeneity of variance across groups. Running a test without checking its assumptions is one of the most common statistical mistakes in student papers, and it is also one of the least visible errors to a student who does not know to look for it.

The fix is specific: run assumption checks before running the primary analysis. For normality, use a Shapiro-Wilk test or inspect a Q-Q plot. For homogeneity of variance, use Levene's test. If assumptions are violated, switch to the appropriate non-parametric alternative. A Wilcoxon signed-rank test replaces a paired t-test when normality is not met. A Spearman correlation replaces a Pearson correlation when the relationship is not linear.

3. Misinterpreting p-values

A p-value below 0.05 does not prove that an effect exists. It means that, if the null hypothesis were true, data as extreme as those observed would occur less than 5% of the time by chance alone. The American Statistical Association issued a formal statement in 2016 clarifying that p-values do not measure the probability that the null hypothesis is true, nor do they measure the size or importance of an effect. Student papers routinely describe a p-value of 0.03 as proof that a treatment worked. This is a misinterpretation that peer reviewers flag consistently.

The correct framing is: the result was statistically significant at the 0.05 level, suggesting the observed difference is unlikely to be due to chance alone. Significance is not synonymous with importance, and a low p-value in a small sample does not generalise to a broader population.

4. Omitting effect sizes

Statistical significance tells a reader whether an effect is likely to be real. Effect size tells a reader how large that effect is. These are separate pieces of information, and both are required for a complete results section. Cohen's d is the standard effect size measure for t-tests. Eta-squared or partial eta-squared is used for ANOVA. For correlations, the r value itself is the effect size. Reporting a p-value without an effect size leaves the reader unable to assess the practical significance of the finding.

5. Inferring causation from correlation

A correlation between two variables means they move together. It does not mean one causes the other. Student papers that use survey data or observational data frequently conclude that variable A caused outcome B when the design of the study can only support the conclusion that A and B are associated. This is not a minor wording issue. It is a fundamental misrepresentation of what the data show, and it is one of the most consequential common statistical mistakes in student papers because it affects the validity of every conclusion in the discussion section.

The fix is to audit every causal claim in the discussion section and replace causal language with associational language. Replace "caused" with "was associated with." Replace "led to" with "correlated with." If the study design is experimental with random assignment, causal language is appropriate. If it is observational or correlational, it is not.

What happens when statistical errors reach peer review?

Peer reviewers assess statistical methodology as a core component of research quality. When a reviewer identifies a statistical error, the paper is not automatically rejected. The most common outcome is a request for major revisions, which requires the author to re-run analyses, rewrite the results section, and resubmit. This adds time to the publication process. At the peer review process stage, reviewers are looking for internal consistency: do the tests chosen match the data type, do the reported statistics match the conclusions drawn, and are the limitations of the design acknowledged honestly?

Desk rejection, which occurs before peer review begins, is less likely to be triggered by statistical errors alone and more likely to occur when the research question is not original or the paper does not meet basic formatting standards. Statistical errors typically surface during full peer review. This means a paper with strong research design and correctable statistical errors has a genuine path to publication after revision. Addressing the five errors above before submission reduces the likelihood of a major revision request and shortens the overall timeline to a decision.

Students who want to see how statistical methods are presented in published work can browse published research to review how authors report tests, effect sizes, and limitations in accepted papers. Reading published methods sections is one of the most efficient ways to calibrate what a complete statistical report looks like.

What are the most common mistakes students make when reporting statistics?

The most common reporting mistakes are: omitting units or sample sizes from tables, rounding p-values to "p = 0.000" instead of reporting "p < 0.001," failing to define abbreviations used in figures, and presenting raw data without descriptive statistics. Each of these errors makes results harder to interpret and signals to reviewers that the analysis was not reviewed carefully before submission.

Reporting p-values as exactly zero is a specific and widespread error. No p-value is exactly zero. When statistical software outputs 0.000, the correct reporting convention is p < 0.001. This is a formatting standard followed by all major journals and is specified in the Publication Manual of the American Psychological Association (APA), 7th edition.

Descriptive statistics must accompany inferential statistics in every results section. Before reporting a t-test result, report the mean and standard deviation for each group. Before reporting a correlation, report the range and distribution of each variable. Reviewers use descriptive statistics to verify that the inferential statistics are plausible. A mean that falls outside the reported range, or a standard deviation larger than the mean for a non-negative variable, signals a data entry or calculation error.

For a broader look at writing errors that affect the clarity of results sections, the guide on common grammar mistakes in academic research papers addresses how imprecise language in results sections obscures otherwise sound findings.

How to audit your paper for statistical errors before submission

1. List every statistical test used in the paper. For each one, confirm that the test matches the data type (categorical or continuous) and the comparison being made (two groups, multiple groups, paired, or independent).

2. Document the assumption checks you ran before each test. If you did not run assumption checks, run them now. Record the results and note which non-parametric alternatives you used where assumptions were violated.

3. Read every sentence in the results section that contains a p-value. Confirm that no sentence describes the p-value as proof of an effect or as the probability that the null hypothesis is true.

4. Confirm that an effect size is reported alongside every significance test. Add Cohen's d, eta-squared, or r where missing.

5. Read every sentence in the discussion section that makes a causal claim. If the study design is observational or correlational, replace causal language with associational language.

6. Check all tables and figures for missing units, undefined abbreviations, and absent sample sizes.

7. Verify that p-values are not reported as exactly zero. Replace any instance of p = 0.000 with p < 0.001.

After completing this audit, review the common mistakes to avoid before preparing your final manuscript. When the paper is ready, review the submission guidelines and submit your work for peer review.

The Princeton Journal of Pre-Collegiate Research publishes original quantitative and qualitative research across all academic disciplines. If your statistical analysis is sound and your paper is ready for peer review, review the submission guidelines at princeton-jpcr.org/submit.

Frequently asked questions about common statistical mistakes in student papers

What is a p-value and what does it actually mean in a student research paper?

A p-value is the probability of observing results at least as extreme as those collected, assuming the null hypothesis is true. It does not measure the probability that the hypothesis is correct or that the result is important. The American Statistical Association's 2016 statement on p-values clarifies that a p-value below 0.05 indicates statistical significance, not practical significance or proof of an effect.

Students should always pair a p-value with an effect size and a clear description of the study's sample size. A statistically significant result in a sample of 15 participants carries far less weight than the same result in a sample of 150.

How long does it take to get feedback on statistical methods after submitting a paper?

The standard peer review and publication timeline at most student journals is 2 to 3 months from submission to a final decision. PJPCR follows this standard timeline. A fast-track option is available for students who need a quicker turnaround, bringing the timeline to 2 to 4 weeks. Submission and peer review are free; a publication fee applies for accepted papers.

Reviewers assess statistical methodology as part of the full review. Papers with significant statistical errors typically receive a request for major revisions rather than an outright rejection, which extends the total timeline.

Do I need a university supervisor to get my statistics right before submitting a paper?

A university supervisor is not required. Many high school students successfully complete statistical analyses using freely available tools such as JASP, which is open-source statistical software designed to be accessible to students without advanced training. The key requirement is that the tests chosen match the data type and that assumption checks are documented.

A school statistics teacher, a science fair mentor, or a publicly available methods textbook can serve as a reference for test selection and assumption checking. What matters is that the analysis is correct and reproducible, not who supervised it.

What makes the statistical analysis in a student paper publishable rather than just adequate?

A publishable statistical analysis does four things: it selects tests appropriate to the data type and design, it documents assumption checks, it reports effect sizes alongside p-values, and it interprets results within the limitations of the study design. An adequate analysis runs a test and reports a p-value. A publishable analysis contextualises that result honestly.

Reviewers also look for transparency about small sample sizes and the resulting limits on generalisability. Acknowledging limitations is not a weakness in a paper. It is a sign of methodological maturity. Students can review the guide on statistical significance for high school researchers for a detailed explanation of how to interpret and report significance correctly.

What kinds of quantitative research does PJPCR publish?

PJPCR publishes original quantitative research across all academic disciplines, including the natural sciences, social sciences, psychology, economics, and interdisciplinary fields. Accepted papers have included experimental studies, survey-based correlational research, and secondary data analyses. The journal does not require university lab access or advanced statistical software as a condition of submission.

All submissions are assessed through the peer review process, which evaluates methodological rigour, originality of the research question, and the accuracy of the conclusions drawn from the data. The journal is selective and does not guarantee acceptance.

Conclusion

The five most consequential common statistical mistakes in student papers are test misselection, skipped assumption checks, p-value misinterpretation, missing effect sizes, and causal overclaiming. Each is correctable before submission. The seven-step audit in this post provides a structured way to identify and fix each error in sequence. Completing that audit before submission reduces the likelihood of a major revision request and produces a results section that reviewers can evaluate with confidence. If your research is ready for peer review after completing that process, submit it to PJPCR at princeton-jpcr.org/submit.