How BioRender recommends the right statistical test for your data

Choosing the right statistical test is one of the most common points of uncertainty in data analysis. BioRender Graphing takes the guesswork out of this by automatically running diagnostic checks on your data and recommending the most appropriate test before you run your analysis. This article explains how that process works, what checks are run, and how to interpret the recommendations.

How the recommendation system works

When you click Run analysis, BioRender runs a series of assumption checks on your data in the background before presenting you with an analysis setup form. Based on the results of those checks, the most appropriate test is marked as "Recommended" in the Review analysis panel on the right-hand side.

The recommendation panel shows:

• The recommended test
• Which checks were run
• What the results mean in plain language
• Whether results from the diagnostic tests will be included in your analysis output

Example below

💡 Important: BioRender’s recommendations are a statistically grounded starting point. You always make the final decision. The reasoning is shown transparently so you can agree with the recommendation, override it, or discuss it with a colleague.

The assumption checks BioRender runs

Different statistical tests have different requirements about the structure and distribution of your data. BioRender automatically checks the most important ones:

Shapiro-Wilk test of normality

Many common statistical tests - including t-tests and ANOVAs - assume that your data follows a normal (bell-curve) distribution. The Shapiro-Wilk test checks whether this assumption holds for your dataset.

Shapiro-Wilk on log-transformed data

If the original data fail the normality check, BioRender automatically tests whether the log-transformed data are normal. Many biological measurements - cytokine concentrations, fluorescence intensities, gene expression - are log-normal by nature, so this second check often "rescues" the data into a parametric framework. When it passes, BioRender recommends the lognormal version of t-test or ANOVA, which runs the parametric test on log-transformed values. This is preferable to the nonparametric alternative because it preserves more information about your effect size.

Levene’s test of equal variances

Parametric tests also assume that the variance across your groups is approximately equal. Levene’s test checks whether this holds.

What you see in the Review Analysis panel

The recommendation appears in the Review analysis panel when you open the analysis setup. 

It shows:

• The recommended test, pre-selected in the dropdown
• A “Checks we ran” section listing which diagnostic tests were performed and whether they passed or failed
• A “What this analysis does” explanation in plain language - for example: “Your data meets the assumptions for an unpaired t-test” or “Your data does not meet the normality assumption for Pearson correlation. Spearman correlation is recommended as a non-parametric alternative.”
• A "When to use" explanation describing the appropriate conditions for the recommended test
• A note confirming that the diagnostic test results will be included in your analysis output

The results of Shapiro-Wilk and Levene’s tests are always included in the full analysis report so you can document the assumption checks in your methods section.

Overriding a recommendation

You are never locked into the recommended test. If you want to use a different test, open the dropdown in the Review analysis panel and select your preferred option.

If you select a test that doesn’t match your data’s assumptions - for example, choosing a parametric test when the normality check fails - BioRender will display a warning explaining why the recommended test is still preferable and what you’re trading off by overriding it. This gives you the information you need to make an informed decision rather than just blocking your choice.

Some tests may appear greyed out if your data structure doesn’t meet their basic requirements. Hover over the tooltip to understand why.

💡 Why this matters for teams: When everyone on a team uses BioRender’s guided analysis, analyses are run consistently - using the same tests, the same assumption checks, and the same logic - regardless of individual statistics knowledge. This reduces variability between lab members and makes methods easier to document and reproduce.

Frequently asked questions

Can I always trust the recommendation? 

The recommendations are based on well-established statistical criteria and are appropriate for the vast majority of common research datasets. However, they are a starting point - you should always consider the conventions of your field and your specific experimental design when making the final decision.

What if my dataset is very small? 

Normality tests like Shapiro-Wilk have limited power with very small samples (e.g., n < 5 per group). In these cases, the test may not detect non-normality even if it exists. Consider the biological plausibility of normality in your data alongside the test result.

Will the assumption check results appear in my analysis output? 

Yes. Shapiro-Wilk and Levene's test results are included in every analysis report so you can report them in your methods section.

What statistical engine does BioRender use? 

All analyses are run using R (version 4.5.1). See the Statistical Analyses in BioRender Graphing (methodology) article for the full list of packages and functions used.

Need help? Reach out to our support team at support@biorender.com or start a live chat by clicking the "Chat With Us" bubble in the bottom right corner.