Use this tool when you have many numeric measurements per sample, for example gene expression, metabolites or clinical biomarkers, and want to see which variables drive the main differences between samples. Upload your data below (CSV, Excel or SPSS, or paste it), optionally add a grouping column, and run the analysis.
About the PCA biplot generator
A PCA biplot places samples (scores) and variables (loadings as arrows) on the same pair of principal components. Principal component analysis (PCA) rewrites many correlated variables as a few uncorrelated components that capture as much variance as possible, so a table with dozens of columns can be inspected in two dimensions.
Arrows that point the same way are positively correlated, arrows at right angles are unrelated, and arrows that point in opposite directions are negatively correlated. A sample sits in the direction of the variables on which it scores high. This generator also checks that PCA is suitable for your data, chooses the number of components with parallel analysis, and writes the results in APA7 style.
When to use it
Use a PCA biplot to explore multivariate data in genomics, metabolomics, microbiome, clinical chemistry, ecology and plant science: to spot batch effects and outliers, to see whether groups separate, to find which variables load together, and to reduce many variables before regression or clustering. It is exploratory: it describes structure, it does not test a hypothesis about groups.
What this generator reports
KMO measure of sampling adequacy and Bartlett’s test of sphericity, to check that the data are suitable for PCA.
Eigenvalues, percentage of variance and cumulative variance for every component, with Horn’s parallel analysis to choose how many to keep.
A table of loadings with salient loadings marked, and the scores of each sample.
A biplot with optional group colours, group ellipses, loading arrows and sample labels, plus a scree plot.
An APA7 results paragraph and table notes, with an interpretation of every table and figure, that you can download as a Word file after logging in.
Assumptions to check
Variables are numeric and roughly linearly related. Strongly skewed variables are better transformed (for example log) first.
There are more samples than variables, and ideally at least 5 to 10 samples per variable.
Variables are on comparable scales or are standardized. Standardizing (a correlation-matrix PCA) stops large-unit variables from dominating.
Missing values are handled before analysis; check outliers, because PCA is sensitive to them.
How to report it in APA7
State the data suitability (KMO and Bartlett’s χ² with degrees of freedom and p), the rule used to retain components (parallel analysis), the number of components kept with their eigenvalues and percentage of variance explained, and the salient loadings (for example |loading| ≥ .40). The generator writes this paragraph for you and lays out the loadings table in APA7 style.
Customize and download
Change colors, fonts, font sizes and line widths in the figure style panel, then download each figure (PNG or SVG) and table (CSV), or the full set of figures, tables and interpretations as one Word file. Downloads need a free StatFas account; running the analysis does not.
Yes. You can analyze data and read all results without an account. Create a free account or log in only when you want to download figures, tables or the Word report.
How many components should I keep?
The tool uses Horn’s parallel analysis: a component is kept when its eigenvalue exceeds the 95th percentile of eigenvalues from random data of the same size. The scree plot shows this comparison.
Can I change colours, fonts and line sizes?
Yes. Figure style controls set the colours, font, font size, line width, marker size and resolution, and every figure and table can be downloaded.
Is my data uploaded to a server?
The analysis runs in your browser. Your data stays on your device while you work with the generator.