In mooreryan/biplotr: Give me data, and I'll give you a snazzy biplot!

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Import needed libraries.

library(ggplot2)
library(biplotr)

Specifying data columns

If you pass a data.frame to pca_biplot that has additional variables (e.g., factors with covariate info) that are not meant to be used in the PCA calculation, you must be sure to specify the data_cols parameter properly.

Let's use a modified version of the iris dataset to see what I mean.

iris2 <- cbind(
  iris, 
  Group = c(
    rep(1, times = nrow(iris) / 2),
    rep(2, times = nrow(iris) / 2)
  )
)

print(str(iris2))

This data, iris2 contains an additional column, Group, that groups the data into two groups, 1 and 2. However, as you can see, it is a numeric column.

class(iris2$Group)

If we are not careful, the Group variable will be included in the PCA calculation even though for our data, it is a covariate that isn't meant to be included.

p <- biplotr::pca_biplot(data = iris2, 
                         point_color = "Species", 
                         arrow_labels = TRUE,
                         limits_nudge_x = 1,
                         limits_nudge_y = -1)
plot(p$biplot)

As you can see, there is an extra variable in the biplot labeled Group. This is because by default, pca_biplot treats every column other than the one specified by point_color as input to the PCA. Since there are more covariates than just the one we specify in point_color, we need to manually specify the data_cols parameter like this:

p <- biplotr::pca_biplot(data = iris2, 
                         data_cols = 1:4,
                         point_color = "Species", 
                         arrow_labels = TRUE,
                         limits_nudge_x = 1,
                         limits_nudge_y = -1)
plot(p$biplot)

What happens if we don't specify data_cols but had specified Group as a factor instead of numeric?

iris2$Group <- as.factor(iris2$Group)

p <- biplotr::pca_biplot(data = iris2, 
                         point_color = "Species", 
                         arrow_labels = TRUE,
                         limits_nudge_x = 1,
                         limits_nudge_y = -1)

In this case, an error occurs with a message somthing like this:

 Error in biplotr::pca_biplot(data = iris2, point_color = "Species", arrow_labels = TRUE, : All columns used for PCA are not numeric! Make sure to select only numeric columns for PCA! (cols included in PCA were: '1, 2, 3, 4, 6')

alerting you that the PCA was attempted on columns 1-4 and 6, which is not what we wanted.

mooreryan/biplotr documentation built on Sept. 2, 2020, 8:32 a.m.