PCA Biplot Explained: How to Make, Read and Report It
A PCA biplot shows your samples and your variables on the same two axes, so you can see at a glance which measurements drive the differences between samples. This guide explains what the plot means, how to check that PCA is suitable, how many components to keep, and how to write the result in APA7, with a free generator you can use in your browser.
What is a PCA biplot?
Principal component analysis (PCA) rewrites many correlated variables as a smaller number of uncorrelated axes, called principal components. The first component captures as much variance as possible, the second captures as much of what is left as possible, and so on. A biplot (Gabriel, 1971) draws the samples as points (their scores) and the variables as arrows (their loadings) on the same pair of components, usually PC1 and PC2.
In genomics, metabolomics, microbiome and clinical research a table can have dozens or thousands of columns. A biplot reduces this to one figure where groups, outliers, batch effects and clusters of related variables become visible (Ringnér, 2008; Jolliffe & Cadima, 2016).
How to read a PCA biplot
- Points close together are samples with similar values across the variables.
- Arrow direction shows which way a variable increases. A sample lies in the direction of the variables on which it scores high.
- Arrow length shows how well the variable is represented on these two components. Short arrows are explained mainly by other components.
- Angle between arrows approximates correlation: a small angle means positive correlation, about 90° means little correlation, about 180° means negative correlation.
- Axis percentages show the share of total variance each component explains. If PC1 and PC2 together explain little, do not over-interpret the picture.
Group ellipses help you judge separation, but remember that PCA is unsupervised: it does not use the group labels, so separation you see is a genuine feature of the variables, not something the method forced.
Check that PCA is suitable: KMO and Bartlett
PCA is only worthwhile when variables are correlated. Two quick checks are standard. The Kaiser–Meyer–Olkin (KMO) measure compares correlations with partial correlations; values above .60 are usually considered acceptable (Kaiser, 1974). Bartlett’s test of sphericity tests whether the correlation matrix differs from an identity matrix; a significant result means there is correlation structure to summarize (Bartlett, 1954).
The PCA Biplot Generator reports both before it draws anything, and flags variables with low measures of sampling adequacy.
How many components should you keep?
The common rules disagree, so pick one and state it. The Kaiser rule (eigenvalue > 1) tends to keep too many components, and the scree plot is subjective. Horn’s parallel analysis (Horn, 1965) is usually the best default: keep a component only if its eigenvalue exceeds the 95th percentile of eigenvalues obtained from random data of the same size. The generator runs this for you and draws the comparison on the scree plot.
For a biplot you still plot two components. Report how much variance they explain and, if parallel analysis keeps more than two, say so and consider showing PC1 vs PC3 as well.
Step by step: make a PCA biplot in StatFas
- Open the PCA Biplot Generator.
- Click Load Example Data, or paste a table from Excel, or upload a CSV, Excel or SPSS file. Rows are samples and columns are numeric variables; add one optional grouping column.
- Choose the numeric variables, the grouping column, the analysed matrix (correlation, i.e. standardized, is recommended when units differ) and the components on the x- and y-axes. Options include 95% group ellipses, sample labels and loading-arrow labels.
- Run the analysis. KMO, Bartlett, eigenvalues, parallel analysis, loadings and scores appear with an APA7 interpretation of every table and figure.
- Use the figure style panel to set colors, font, font size, line width and marker size, then log in (free) to download each figure as PNG or SVG, each table as CSV, or everything as a Word file.
Prefer code? The step-by-step tutorial shows the same plot in StatFas, SPSS, R, Python and MATLAB.
How to report a PCA biplot in APA7
Report the data suitability checks, the retention rule, the number of components with their eigenvalues and percentage of variance, and the salient loadings (commonly |loading| ≥ .40). Include the biplot as a numbered figure with a note about scaling and the groups shown. Example wording (illustrative numbers only):
Common mistakes to avoid
- Not standardizing variables measured in different units, so the largest-variance variable dominates PC1.
- Reading causes into arrows. A biplot shows association, not mechanism.
- Ignoring explained variance. A biplot of two components that explain 20% of the variance is a weak summary.
- Using PCA on very few samples relative to variables, which gives unstable loadings.
- Over-reading separation caused by a batch or run order effect rather than biology.
Frequently asked questions
What is the difference between PCA and factor analysis?
PCA summarizes total variance with components that are weighted sums of the variables. Factor analysis models shared variance caused by latent factors. PCA is a data reduction method; use exploratory factor analysis when you want to model latent constructs.
Should I standardize my variables before PCA?
Yes when variables have different units or very different variances. Standardizing makes PCA run on the correlation matrix so each variable contributes equally.
How many samples do I need for PCA?
There is no single rule, but at least 5 to 10 samples per variable and more samples than variables are common guidelines. Fewer samples make loadings unstable.
Can I make a PCA biplot for free without coding?
Yes. The StatFas PCA Biplot Generator is free to run in your browser with no account; a free account is needed only to download results.
Why do my PCA results differ between R and SPSS?
Differences usually come from scaling (covariance vs correlation matrix), the sign of components (arbitrary), rotation, or how missing values are handled. Check each program’s settings.
Related tools, guides and tutorials
- Tutorial: PCA biplot in StatFas, SPSS, R and Python
- PCA calculator
- PCA tutorial
- Principal component analysis explained
- Exploratory factor analysis
- EFA tutorial
- Heatmap maker
- Hypothesis and analysis suggester
- Parallel Coordinates Plot Generator
- 3D Scatter Plot Generator
- Ternary Plot Generator
- All tutorials
- All blog posts
References
- American Psychological Association. (2020). Publication manual of the American Psychological Association (7th ed.).
- Bartlett, M. S. (1954). A note on the multiplying factors for various χ² approximations. Journal of the Royal Statistical Society B, 16(2), 296–298.
- Gabriel, K. R. (1971). The biplot graphic display of matrices with application to principal component analysis. Biometrika, 58(3), 453–467.
- Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30(2), 179–185.
- Jolliffe, I. T., & Cadima, J. (2016). Principal component analysis: A review and recent developments. Philosophical Transactions of the Royal Society A, 374(2065), 20150202.
- Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, 39(1), 31–36.
- Ringnér, M. (2008). What is principal component analysis? Nature Biotechnology, 26(3), 303–304.