Implementation of corrplot using ggplot2
Reinventing wheels is not what I like doing. corrplot is a great R package, but I am really tired of customizing the appearance of corrplot, for example, the space between colorbar and its tick labels, the space around the plot that I don’t know how to control when writing it to PDF on my macOS. This is most likely because I am more familiar with the Grammar of Graphics implemented in ggplot2 than the base plotting system in R. There are several R packages (e.g., ggcorrplot developed by Alboukadel Kassambara, ggcorr developed by François Briatte) that can visualize a correlation matrix into a corrgram using ggplot2; however, they are unable to visualize a correlation matrix using ellipse and mixed methods. ggcorrplot2 has implemented only a subset of features of corrplot to meet my urgent needs. See examples in the Getting started section. More functionality will be added in the future as needed.
Get the development version from github:
if (!requireNamespace("devtools")) install.packages("devtools")
devtools::install_github("caijun/ggcorrplot2")The mtcars dataset will be used to demonstrate the usages of
ggcorrplot2. Most parameters of ggcorrplot2 functions are the
same as those of corrplot. Therefore, it’s easy for users to migrate
from corrplot to ggcorrplot2.
library(ggcorrplot2)
data(mtcars)
# Use corr.test() from psych package to calculate the correlation matrix and
# corresponding p value matrix without adjustment.
library(psych)
ct <- corr.test(mtcars, adjust = "none")
corr <- ct$r
p.mat <- ct$p
# Visualize the correlation matrix
# --------------------------------
# method = "circle" (default)
ggcorrplot(corr)
# method = "square"
ggcorrplot(corr, method = "square")
# method = "ellipse"
ggcorrplot(corr, method = "ellipse")
# method = "number", display the correlation coefficients
ggcorrplot(corr, method = "number")
# Visualize the upper or lower triangle of correlation matrix
# -----------------------------------------------------------
# the upper triangle
ggcorrplot(corr, type = "upper")
# the lower triangle
ggcorrplot(corr, type = "lower")
# Visualize the correlation matrix using mixed methods
# ----------------------------------------------------
# default: upper = "circle", lower = "number"
ggcorrplot.mixed(corr)
# upper = "ellipse", lower = "number"
ggcorrplot.mixed(corr, upper = "ellipse", lower = "number")
# Combine correlogram with the significance test
# ----------------------------------------------
# Insignificant coefficients according to the default significant level
# (sig.lvl = 0.05) are indicated by X by default.
ggcorrplot.mixed(corr, upper = "ellipse", lower = "number", p.mat = p.mat)
# Leave blank on insignificant coefficients
ggcorrplot.mixed(corr, upper = "ellipse", lower = "number", p.mat = p.mat,
insig = "blank")
# Label significant coefficients with asterisks (*, default) denoting the significance level
ggcorrplot.mixed(corr, upper = "ellipse", lower = "number", p.mat = p.mat,
insig = "label_sig", sig.lvl = c(0.05, 0.01, 0.001))
# Label significant coefficients with varying number of + denoting the significance level
(p <- ggcorrplot.mixed(corr, upper = "ellipse", lower = "number", p.mat = p.mat,
insig = "label_sig", sig.lvl = c(0.05, 0.01, 0.001), pch = "+",
pch.cex = 4))
The above examples reproduce some features of corrplot. In the following example, the added advantages of implementing corrplot using ggplot2, such as customizing the appearance of corrgram, combining a corrgram with other plots (including non-corrgrams) into one plot using cowplot, are demonstrated.
# Customize the appearance of corrplot using functions from ggplot2
library(ggplot2)
# Use different color palette
col1 <- colorRampPalette(c("#7F0000", "red", "#FF7F00", "yellow", "white",
"cyan", "#007FFF", "blue", "#00007F"))
# Change the colorbar direction to horizontal and place it at the bottom
# As mixed methods are used, there are two scales: color filled in ellipse and
# number color
p <- p + scale_fill_gradientn(colours = col1(10), limits = c(-1, 1),
guide = guide_colorbar(
direction = "horizontal",
title = "",
nbin = 1000,
ticks.colour = "black",
frame.colour = "black",
barwidth = 15,
barheight = 1.5)) +
scale_colour_gradientn(colours = col1(10), limits = c(-1, 1),
guide = guide_colorbar(
direction = "horizontal",
title = "",
nbin = 1000,
ticks.colour = "black",
frame.colour = "black",
barwidth = 15,
barheight = 1.5)) +
theme(legend.position = "bottom")
p
# Combine a lower corrgram and a mixed corrgram side by side with a shared colorbar on the bottom
# NAs are allowed in correlation matrix and p value matrix, which are labelled as NA
# Assign NAs
rid <- c(2, 3, 1, 2)
cid <- c(3, 2, 2, 1)
pos <- cbind(rid, cid)
corr[pos] <- NA
p.mat[pos] <- NA
# a lower corrgram
p1 <- ggcorrplot(corr, type = "lower", method = "square", p.mat = p.mat,
insig = "label_sig", sig.lvl = c(0.05, 0.01, 0.001), show.diag = FALSE)
# a mixed corrgram
p2 <- ggcorrplot.mixed(corr, upper = "ellipse", lower = "number", p.mat = p.mat,
insig = "label_sig", sig.lvl = c(0.05, 0.01, 0.001),
pch = "+", pch.cex = 4)
library(cowplot)
prow <- plot_grid(p1 + theme(legend.position = "none"),
p2 + theme(legend.position = "none"),
rel_widths = c(1, 1), nrow = 1, align = 'hv',
labels = c("(a)", "(b)"), label_x = 0, label_y = 1)
# Extract the legend from the first corrgram
legend <- get_legend(p1)
# Add the legend to the bottom of the plot row we made earlier.
p <- cowplot::plot_grid(prow, legend, ncol = 1, rel_heights = c(1, 0.15))
p
To cite the ‘ggcorrplot2’ package in publications use:
Jun Cai, Granville Matheson and Samson Leonard Daniël (2022). ggcorrplot2: Visualize a Correlation Matrix using ggplot2. R package version 0.1.2.
A BibTeX entry for LaTeX users is
@Manual{,
title = {ggcorrplot2: Visualize a Correlation Matrix using ggplot2},
author = {Jun Cai and Granville Matheson and Samson Leonard Daniël},
year = {2022},
note = {R package version 0.1.2},
url = {https://github.com/caijun/ggcorrplot2},
}
The above citation information can be generated by calling
citation("ggcorrplot2") in R.
Bugs and feature requests can be filed to https://github.com/caijun/ggcorrplot2/issues. Pull requests are also welcome.