Plot upper triangle correlation matrix with similarity scores using ggplot

Question

I have a dataframe as given below:

The table only has values from the upper triangle of a matrix. I want to plot a correlation plot (correlogram) where the colours show the correlation and size according to the similarity scores. I tried plotting using ggplot2:

ggplot(df, aes(x = var1, y = var2, color = cor, size = jaccard)) +
geom_point(aes(), alpha = 0.7) +
scale_size_continuous(name = "Correlation") +
scale_color_continuous(name = "Jaccard Index", low = "blue", high = "red")

When I use the above code, it plots the entire matrix and the plotting is scattered (sample below).

I want to make a neat plot with values showing just the upper triangle. How to do this in R?

Share your dataframe fully dput(mydata), or show us the codes how you created that dataframe. — zx8754
– zx8754, Commented Apr 29, 2024 at 11:01
We can't do anything with a picture of your data, post the output of dput(mydata), so we have the same reproducible data. — zx8754
– zx8754, Commented Apr 29, 2024 at 11:27
@KamalikaRay I have figured out the issue with your chart. To fix it, all you need to do is convert var1 and var2 into factors along with the desired order of the levels specified. Below, I have replicated the problem using the "mtcars" dataset. One thing which I couldn't figure out was how to calculate "jaccard index," and thus, I decided to use dummy values. I would appreciate it if you could share how to actually calculate "jaccard index." Thank you! — StatisticsFanBoy
– StatisticsFanBoy, Commented Apr 30, 2024 at 18:22

StatisticsFanBoy · Accepted Answer · 2024-05-01 01:24:46Z

I have simulated your problem using the "mtcars" dataset. See code below.

install.packages(c("tidyverse", "foreach"))
data(mtcars)

data <- mtcars
colnames(data) <- paste0("var", 1:length(data)) # rename column names as var1, ..., var11 (A, ..., L in pictured data frame)

newdata <- data.frame(column_one = rep(colnames(data)[1:length(data)-1], times = seq(from = length(data)-1, to = 1, by = -1))) # create column 1 of the dataset (var1 in the pictured data frame)

library(foreach)
newdata$column_two <- foreach(i = 2:length(data), .combine="c") %do% {
rep(colnames(data)[i:length(data)], each=1)
} # create column 2 of the dataset (var2 in the pictured data frame)

newdata$column_three <- foreach(i = newdata$column_one, j = newdata$column_two, .combine = "c") %do% {
  cor(data[[i]], data[[j]])
} # calculate correlations and create column 3 of the dataset (correlation in the pictured data frame)

jaccard <- function(a, b) {
  intersection = length(intersect(a, b))
  union = length(a) + length(b) - intersection
  return (intersection/union)
} # https://www.r-bloggers.com/2021/11/how-to-calculate-jaccard-similarity-in-r-2/

newdata$column_four <- foreach(i = newdata$column_one, j = newdata$column_two, .combine = "c") %do% {
  jaccard(data[[i]], data[[j]])
} # calculate jaccard index (jaccard in the pictured data frame)

lapply(newdata, class) # column_one, column_two should be character vectors
# column_three, column_four should be numeric vectors

# if the outcome of lapply() is otherwise, run the below four lines
newdata$column_one <- as.character(newdata$column_one)
newdata$column_two <- as.character(newdata$column_two)
newdata$column_three <- as.numeric(newdata$column_three)
newdata$column_four <- as.numeric(newdata$column_four)

newdata$column_one <- factor(newdata$column_one, levels = c("var1", "var2", "var3", "var4", "var5", "var6", "var7", "var8", "var9", "var10", "var11")) # convert column one into factor with the desired order of the levels specified
newdata$column_two <- factor(newdata$column_two, levels = c("var2", "var3", "var4", "var5", "var6", "var7", "var8", "var9", "var10", "var11")) # convert column two into factor with the desired order of the levels specified

library(tidyverse)
ggplot(newdata, aes(x = column_one, y = column_two, color = column_three, size = column_four)) +
  geom_point(aes(), alpha = 0.7) +
  scale_size_continuous(name = "Jaccard Index") +
  scale_color_continuous(name = "Correlation", low = "blue", high = "red")

The final bubble chart will look like this:

Additional comment: I noticed a minor issue with your code. You specified plot colors based on correlations (color = cor) but your color scale is labelled as "jaccard index" (scale_color_continuous(name = "Jaccard Index", low = "blue", high = "red"))

Collectives™ on Stack Overflow

Plot upper triangle correlation matrix with similarity scores using ggplot

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related