PCA score plot with group ellipses
Compare three penguin species using four body measurements.
Source: Palmer penguins (Gorman et al., 2014)
PCA Plot Maker
FigranPaste a data table or PC scores and get a publication-ready PCA plot.
- Score plot, biplot and scree plot
- From raw data or your scores
- Free SVG and 300 DPI PNG
PCA plot examples
Explore real datasets and load a PCA plot example to make it your own.
Biplot of iris measurements
See how four flower measurements relate to species.
Source: R iris data
PCA biplot with sample labels
Compare state-level arrest statistics with labeled observations.
Source: R USArrests data
Scree plot with cumulative variance
Examine the variance explained by wine measurements.
Source: UCI Wine dataset
Correlation circle (loadings plot)
Explore associations among standardized wine measurements.
Source: UCI Wine dataset
Gene expression PCA
Compare cancer cell lines using the 500 most variable genes.
Source: ISLR2 NCI-60 dataset
Scores from R prcomp
Plot precomputed cancer cell line coordinates on PC1 and PC3.
Source: ISLR2 NCI-60 dataset
UMAP plot of handwritten digits
Plot any 2-D embedding from scores, including 1,797 handwritten digits.
Source: UCI digits; UMAP (seed 42)
How to make a PCA plot
Make a principal component analysis plot from a table in four steps.
1Enter data
Paste a sample-by-variable table or existing coordinates for your PCA plot.
2Set PCA options
Choose sample direction and whether to scale variables.
3Choose a plot
Select scores, a biplot or a scree plot.
4Download
Save PNG or SVG and take the results and figure legend with you.
How to read a PCA plot
A PCA plot summarizes patterns in many variables on a small number of axes.
Score plot
To interpret a PCA plot, start with the points and axis percentages. Each point represents a sample, and the percentages show how much total variation each component explains. Samples close together have similar profiles in the displayed components; a distant point may deserve a closer look.
Colors and ellipses help compare groups, but an ellipse describes spread rather than proving that groups differ.
- 1Axis percentages
- 2Group separation
- 3Potential outlier
- 4Variable arrow
Biplot and loadings plot
Arrows show how variables relate to the displayed components. Long arrows contribute strongly in this view, and arrows pointing in similar directions indicate similar patterns. In a scaled PCA loadings plot, the correlation circle helps interpret variable correlations.
Axis aspect ratio matters when comparing directions and distances (Nguyen and Holmes, 2019).
Scree plot and explained variance
Use the shape of the curve and the cumulative total together when choosing components.
A scree plot orders components by explained variance. An elbow marks where the gain from further components starts to level off, but the choice depends on the analysis and should not rely on an automatic cutoff (Cattell, 1966).
For standardized variables an eigenvalue of one is sometimes used as a reference; it is not a universal decision rule. The cumulative line shows how much variation the displayed components capture together.
| PC | % explained | Cumulative % |
|---|---|---|
| PC1 | 36.2% | 36.2% |
| PC2 | 19.2% | 55.4% |
| PC3 | 11.1% | 66.5% |
| PC4 | 7.1% | 73.6% |
PCA plot FAQ
How do I make a PCA plot?
Paste a sample-by-variable table into Data table to calculate PCA, or choose Scores and paste coordinates calculated elsewhere. Select a plot and download the result.
Should I scale my data before PCA?
Scale variables when their units or magnitudes differ. When variables share units and their variances are meaningful, such as variance-stabilized expression values, leaving them unscaled may be appropriate.
What does the percentage on each PCA axis mean?
It is the proportion of total variance explained by that principal component. The first component explains the most variation, followed by the second.
How many principal components should I keep?
Use the scree plot and cumulative explained variance alongside the purpose of your analysis. An elbow can suggest a useful range, but there is no universal threshold.
What is a biplot, and how do I read the arrows?
A PCA biplot combines sample scores and variable directions. Arrows point toward increasing values; their direction and length show their relationship to the displayed components. The figure legend records the arrow scaling used here.
Why are my PCA signs flipped compared with R or Python?
Principal component signs are arbitrary, so a mirrored axis represents the same solution. Use Flip PC1 or Flip PC2 to match a reference figure.
Can I plot UMAP or t-SNE coordinates?
Yes. Paste existing coordinates in Scores mode. UMAP and t-SNE are different from PCA: their axes do not have PCA explained-variance percentages, and distances require careful interpretation.
Is my data uploaded? What can I download?
Your data stays in your browser. Download SVG, 300 DPI PNG and results CSV files or copy a figure legend, without signing in.
Make your PCA plot
Turn your data into a PCA plot for your next paper or presentation.
Make your PCA plot