CODEWITHSUNDEEP
rggplot2kernel-density
3N4N
3N4N

Reputation: 619

use ggplot2 with kde plots of ks package

I'm using ks package to calculate and plot bivariate kernel density estimations. It works great, but I want to use ggplot to plot the estimations instead.

I could try and create a custom function to prepare a data frame and stuff for ggplot but I have a feeling more experienced users of R could do this with less effort.

So, how do I do it?

Here is an example scenario of ks::kde:

data <- cbind(rnorm(100), rnorm(100))
kd <- ks::kde(data)
plot(kd, display = "slice", col = viridis::viridis(20))
plot(kd, display = "image", col = viridis::viridis(20))
plot(kd, display = "persp", col.fun = viridis::viridis)

The above code should give you the following plots. How do I use ggplot to achive these plots?

EDIT: I need to use the estimations calculated by ks package. The native functions provided by ggplot2 uses a package named MASS. So that won't do.

slice plot image plot perspective plot

Upvotes: 0

Views: 1147

Answers (1)

Allan Cameron
Allan Cameron

Reputation: 173793

The first two are straightforward in ggplot. Here's a fully reproducible example:

library(ggplot2)

data <- cbind(rnorm(100), rnorm(100))

d <- ggplot(as.data.frame(data), aes(V1, V2)) + 
  scale_color_viridis_c() +
  theme_bw()

d + geom_density2d(aes(colour = after_stat(level)), bins = 5)


d + geom_density2d_filled()

EDIT

To specifically use the output of ks::kde(), we need to convert its data from a matrix into long format using reshape2::melt, but then it is still very easy to plot in ggplot:

set.seed(1)

data <- cbind(rnorm(100), rnorm(100))
kd   <- ks::kde(data)

library(ggplot2)

Contour Plot

ggplot(reshape2::melt(kd$estimate)) +
  geom_contour(aes(x = Var1, y = Var2, z = value, color = stat(level)),
               bins = 5) +
  scale_color_viridis_c() +
  coord_equal() +
  theme_classic()

Filled contour plot

ggplot(reshape2::melt(kd$estimate)) +
  geom_contour_filled(aes(x = Var1, y = Var2, z = value, fill = stat(level)),
               bins = 9) +
  scale_fill_viridis_d() +
  coord_equal() +
  theme_classic() +
  theme(legend.position = "none")

Raster image

ggplot(reshape2::melt(kd$estimate)) +
  geom_raster(aes(x = Var1, y = Var2, fill = value)) +
  scale_fill_viridis_c() +
  coord_equal() +
  theme_classic() +
  theme(legend.position = "none")

Created on 2021-11-12 by the reprex package (v2.0.0)

Upvotes: 5

Related Questions