Summary
Group comparisons were performed according to data distribution (Shapiro–Wilk test) and variance assumptions (Levene’s test).
For two-group comparisons, normally distributed data were tested using Student’s t-test (equal variances) or Welch’s t-test (unequal variances), based on Levene’s test or user selection. Non-parametric data were evaluated using the Wilcoxon rank-sum test.
For three or more groups, normally distributed data were analyzed using one-way ANOVA (equal variances) or Welch’s ANOVA (unequal variances), based on Levene’s test or user selection. Non-parametric distributions were compared using the Kruskal–Wallis test.
When overall group differences were significant, pairwise comparisons were performed using:
- Student’s t-tests with Holm correction (following one-way ANOVA),
- Games–Howell tests (following Welch’s ANOVA), or
- Dunn’s tests with Holm correction (following Kruskal–Wallis).
Holm–Bonferroni correction is set as the default, but you can also select the Bonferroni correction.
The underlying concept builds on statistical procedures used in my previous publications, incorporating modifications and enhancements for broader applicability [1-3].
This tool is built with R.
How to use
Data input
Upload the Excel file that contains your dataset and specify the sheet name to be used for the analysis (it automatically selects the name of first sheet).
The first row must include variable names in the excel file, and one of the columns must indicate group identity (e.g., Group, Sample, Category).
Alternative Input: “Edit / Paste Table”
You may also enter data by using the Edit / Paste Table option.
For best results: Use the first column as an identifier (e.g., Group 1,2,3).
All other columns should contain numeric values.
You do not need to paste the header row. If you do paste it, the non-numeric text will be automatically removed and may not affect the analysis.
Run the analysis
Click “Start group comparison” to generate:
Summary statistics (e.g., mean ± SD or median (IQR), P-values)
Interactive plots (for selected variable)
Bars indicate mean values for parametric datasets or median values for non-parametric datasets.
*This method considers the family-wise error rate (FWER) but does not apply false discovery rate (FDR) corrections.
*If you don’t know whether variance assumptions are satisfied or not in your datasets, please use Welch’s t-test. Variance tests such as Levene’s test are often unreliable with small or unbalanced samples. It is very helpful to read this: How do you know if the standard deviation of the groups is the same? (Statistics Kingdom)
References
[1] Cseh, Domonkos, et al. “Phenotypes of Hypertension: Impact of Age and Sex on Hemodynamic Mechanisms.” Journal of the American Heart Association 14.17 (2025): e042096.
[2] Kamel, Rima, et al. “Cardiac Gene Therapy With Phosphodiesterase 2A Limits Remodeling and Arrhythmias in Mouse Models of Heart Failure.” Journal of the American Heart Association 14.3 (2025): e037343.
[3] Ryu, Jae Kyu, et al. “Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration.” Nature immunology 19.11 (2018): 1212-1223.