PCA Biplot in R with ggfortify and factoextra

Running PCA

Use prcomp() to run PCA on the numeric columns of your data. Set scale. = TRUE to standardize variables to the same scale — essential when variables have different units or magnitudes.

pca <- prcomp(iris[, 1:4], scale. = TRUE)

Base R: `biplot()`

The quickest approach is base R. biplot() overlays the observations (row labels) and variables (arrows) on the same axes.

pca <- prcomp(iris[, 1:4], scale. = TRUE)

biplot(pca)

PCA biplot in base R with biplot function

`ggfortify::autoplot()`

PCA biplot in ggplot2 with ggfortify autoplot and variable loadings

ggfortify adds a prcomp method to ggplot2’s autoplot(). Set loadings = TRUE to draw the variable arrows and loadings.label = TRUE to label them.

# install.packages("ggfortify")
library(ggfortify)

autoplot(pca,
         loadings       = TRUE,
         loadings.label = TRUE)

Color by group

Pass the original data frame to data and map a grouping column to colour to color the observations. The loadings arrows are drawn on top.

# install.packages("ggfortify")
library(ggfortify)

autoplot(pca,
         data           = iris,
         colour         = "Species",
         loadings       = TRUE,
         loadings.label = TRUE)

PCA biplot with ggfortify colored by iris species

`factoextra::fviz_pca_biplot()`

PCA biplot with factoextra fviz_pca_biplot

factoextra is the most popular package for PCA visualization. fviz_pca_biplot() produces a polished ggplot2 biplot out of the box with labeled arrows and variance explained on the axes.

# install.packages("factoextra")
library(factoextra)

fviz_pca_biplot(pca)

Groups and confidence ellipses

Pass a grouping factor to habillage and set addEllipses = TRUE to draw confidence ellipses around each group. This is the most common way to present PCA results in publications.

# install.packages("factoextra")
library(factoextra)

fviz_pca_biplot(pca,
                habillage   = iris$Species,
                addEllipses = TRUE)

PCA biplot with confidence ellipses per group using factoextra

Variable plot

PCA variable plot colored by contribution with factoextra fviz_pca_var

fviz_pca_var() shows only the variable arrows. Set col.var = "contrib" to color each variable by its percentage contribution to the principal components.

# install.packages("factoextra")
library(factoextra)

fviz_pca_var(pca,
             col.var       = "contrib",
             gradient.cols = c("#00AFBB", "#E7B800", "#FC4E07"))