PCA Biplot with ggplot2 — ggplot

Produces a ggplot2 variant of a so-called biplot for PCA (principal component analysis), but is more flexible and more appealing than the base R biplot() function.

ggplot_pca(
  x,
  choices = 1:2,
  scale = 1,
  pc.biplot = TRUE,
  labels = NULL,
  labels_textsize = 3,
  labels_text_placement = 1.5,
  groups = NULL,
  ellipse = TRUE,
  ellipse_prob = 0.68,
  ellipse_size = 0.5,
  ellipse_alpha = 0.5,
  points_size = 2,
  points_alpha = 0.25,
  arrows = TRUE,
  arrows_colour = "darkblue",
  arrows_size = 0.5,
  arrows_textsize = 3,
  arrows_textangled = TRUE,
  arrows_alpha = 0.75,
  base_textsize = 10,
  ...
)

Arguments

x	an object returned by `pca()`, `prcomp()` or `princomp()`
choices	length 2 vector specifying the components to plot. Only the default is a biplot in the strict sense.
scale	The variables are scaled by `lambda ^ scale` and the observations are scaled by `lambda ^ (1-scale)` where `lambda` are the singular values as computed by `princomp`. Normally `0 <= scale <= 1`, and a warning will be issued if the specified `scale` is outside this range.
pc.biplot	If true, use what Gabriel (1971) refers to as a "principal component biplot", with `lambda = 1` and observations scaled up by sqrt(n) and variables scaled down by sqrt(n). Then inner products between variables approximate covariances and distances between observations approximate Mahalanobis distance.
labels	an optional vector of labels for the observations. If set, the labels will be placed below their respective points. When using the `pca()` function as input for `x`, this will be determined automatically based on the attribute `non_numeric_cols`, see `pca()`.
labels_textsize	the size of the text used for the labels
labels_text_placement	adjustment factor the placement of the variable names (`>=1` means further away from the arrow head)
groups	an optional vector of groups for the labels, with the same length as `labels`. If set, the points and labels will be coloured according to these groups. When using the `pca()` function as input for `x`, this will be determined automatically based on the attribute `non_numeric_cols`, see `pca()`.
ellipse	a logical to indicate whether a normal data ellipse should be drawn for each group (set with `groups`)
ellipse_prob	statistical size of the ellipse in normal probability
ellipse_size	the size of the ellipse line
ellipse_alpha	the alpha (transparency) of the ellipse line
points_size	the size of the points
points_alpha	the alpha (transparency) of the points
arrows	a logical to indicate whether arrows should be drawn
arrows_colour	the colour of the arrow and their text
arrows_size	the size (thickness) of the arrow lines
arrows_textsize	the size of the text at the end of the arrows
arrows_textangled	a logical whether the text at the end of the arrows should be angled
arrows_alpha	the alpha (transparency) of the arrows and their text
base_textsize	the text size for all plot elements except the labels and arrows
...	arguments passed on to functions

Source

The ggplot_pca() function is based on the ggbiplot() function from the ggbiplot package by Vince Vu, as found on GitHub: https://github.com/vqv/ggbiplot (retrieved: 2 March 2020, their latest commit: 7325e88; 12 February 2015).

As per their GPL-2 licence that demands documentation of code changes, the changes made based on the source code were:

Rewritten code to remove the dependency on packages plyr, scales and grid
Parametrised more options, like arrow and ellipse settings
Hardened all input possibilities by defining the exact type of user input for every argument
Added total amount of explained variance as a caption in the plot
Cleaned all syntax based on the lintr package, fixed grammatical errors and added integrity checks
Updated documentation

Details

The colours for labels and points can be changed by adding another scale layer for colour, such as scale_colour_viridis_d() and scale_colour_brewer().

Stable Lifecycle

The lifecycle of this function is stable. In a stable function, major changes are unlikely. This means that the unlying code will generally evolve by adding new arguments; removing arguments or changing the meaning of existing arguments will be avoided.

If the unlying code needs breaking changes, they will occur gradually. For example, a argument will be deprecated and first continue to work, but will emit an message informing you of the change. Next, typically after at least one newly released version on CRAN, the message will be transformed to an error.

Examples

# `example_isolates` is a data set available in the AMR package.
# See ?example_isolates.

# See ?pca for more info about Principal Component Analysis (PCA).
if (require("dplyr")) {
  pca_model <- example_isolates %>% 
    filter(mo_genus(mo) == "Staphylococcus") %>% 
    group_by(species = mo_shortname(mo)) %>%
    summarise_if (is.rsi, resistance) %>%
    pca(FLC, AMC, CXM, GEN, TOB, TMP, SXT, CIP, TEC, TCY, ERY)
    
  # old (base R)
  biplot(pca_model)
  
  # new 
  ggplot_pca(pca_model)
  
  if (require("ggplot2")) {
    ggplot_pca(pca_model) +
      scale_colour_viridis_d() +
      labs(title = "Title here")
  }
}