R Tutorial - Visualization of correlation matrix in R using ggcorrplot package in R - ggcorrplot tutorial - ggplot2 extension

ggcorrplot.R

# Demonstration of ggcorrplot package 
# Package created by Alboukadel Kassambara
# Reference link - https://cran.r-project.org/web/packages/ggcorrplot/ggcorrplot.pdf


# Load the required libraries
# install.packages("ggcorrplot")
library(ggcorrplot)

# mtcars dataset from base R used here for demonstration
data(mtcars)

# Structure of mtcars dataset
str(mtcars)

# help page of dataset
# ?mtcars 

######### Correlation matrix and visualization #####
# Step - 1 : Calculate the correlation measure
cor(mtcars)
cor = round(cor(mtcars), 1)

# Step - 2 : Applying the ggcorrplot() on correlation matrix just created
ggcorrplot(cor) # square tiles by default

# Applying the ggcorrplot with method="circle"
ggcorrplot(cor, method = "circle")

# Applying hc order to it
# Correlation matrix will be ordered based on hierarchical clustering method
ggcorrplot(cor, 
           method = "circle", 
           hc.order = TRUE)

# with square tiles
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE)


# Correlogram layout - type = "lower"
ggcorrplot(cor, 
           method = "circle", 
           hc.order = TRUE, type = "lower")

ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "lower")

# Correlogram layout - type = "upper"
ggcorrplot(cor, 
           method = "circle", 
           hc.order = TRUE, type = "upper")

######### Formatting the Correlation matrix #######
## Add a title to correlogram
ggcorrplot(cor, 
           method = "circle", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset")

## Add a legend title
ggcorrplot(cor, 
           method = "circle", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr" 
             )

## Add correlation coefficients
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr", lab = TRUE
)

# Change the coefficient label size and color
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr", 
           lab = TRUE, lab_col ="white" , lab_size = 2
)

## Add a ggplot2 theme 
# theme_light
# theme_void
# theme_dark
# theme_grey
# theme_linedraw

library(ggplot2)
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr", 
           lab = TRUE, lab_size = 2, ggtheme = theme_void
)


# add outline color to tiles
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr", 
           lab = TRUE, lab_size = 2, 
           ggtheme = theme_void, outline.color = "black"
)


# change the legend colors
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, type = "upper", 
           title = "Correlogram - mtcars dataset", 
           legend.title = "Pearson \n Corr", 
           lab = TRUE, lab_size = 2, 
           ggtheme = theme_void, 
           outline.color = "white",
           colors = c("#6D9EC1", "white", "#E46726")
)


######### Calculating p values and marking/dropping insignificant correlation coefficients  ######

# Calculate p-values of the correlation for significance level
p.mat = cor_pmat(mtcars)

ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, 
           outline.color = "white",
           type = "upper",
           p.mat = p.mat
)

# If the p-value in p-mat is bigger than sig.level, then the corresponding correlation coefficient is regarded as insignificant
# default significance level is 0.05 but can be changed

## Blank to insignificant levels
ggcorrplot(cor, 
           method = "square", 
           hc.order = TRUE, 
           outline.color = "white",type = "upper",
           p.mat = p.mat, insig = "blank"
)