r - How to set alpha by bivariate density and fill by z in geom_raster? - Stack Overflow

admin管理员组

文章数量:1343687

I make a heatplot with geom_raster, something like this:

mround <- function(x,base){base*round(x/base)} # round to nearest 5
df <- data.frame(x= mround(rnorm(10000, 100, 10), 5),
                 y= mround(rnorm(10000, 100, 10), 5))
df$z <- df$x + df$y
library(ggplot2)
ggplot(df) +
  geom_raster(aes(x= x, y= y, fill= z)) +
  theme_classic()

I would like to set the alpha argument for the raster and it should depent on the bivariate density. We can visualize the bivariate density for raster like this:

ggplot() +
  stat_density_2d(data= df, aes(x= x, y= y, alpha = after_stat(density)), geom = "raster", contour= FALSE)

Basically, I want the first plot to have the alpha of the second. I tried all kind of thing, like geom_raster(aes(x= x, y= y, fill= z, alpha = after_stat(density))), but nothing works so far. How can I set the fill by z and alpha by 2d density at the same time?

I make a heatplot with geom_raster, something like this:

mround <- function(x,base){base*round(x/base)} # round to nearest 5
df <- data.frame(x= mround(rnorm(10000, 100, 10), 5),
                 y= mround(rnorm(10000, 100, 10), 5))
df$z <- df$x + df$y
library(ggplot2)
ggplot(df) +
  geom_raster(aes(x= x, y= y, fill= z)) +
  theme_classic()

I would like to set the alpha argument for the raster and it should depent on the bivariate density. We can visualize the bivariate density for raster like this:

ggplot() +
  stat_density_2d(data= df, aes(x= x, y= y, alpha = after_stat(density)), geom = "raster", contour= FALSE)

Basically, I want the first plot to have the alpha of the second. I tried all kind of thing, like geom_raster(aes(x= x, y= y, fill= z, alpha = after_stat(density))), but nothing works so far. How can I set the fill by z and alpha by 2d density at the same time?

• r
• ggplot2
• fill
• alpha
• kernel-density

Share Improve this question Follow edited Nov 19, 2024 at 14:52 LulY asked Nov 19, 2024 at 14:45 LulYLulY 1,30422 gold badges1212 silver badges2727 bronze badges 2

• I think this code cannot be executed. Looks faulty. You are missing library calls. Where is mround from? – Friede Commented Nov 19, 2024 at 14:48
• 1 @Friede Edited. – LulY Commented Nov 19, 2024 at 14:52

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 3

This is difficult, because in order to calculate density, ggplot will have to aggregate your data, which loses the value of variables other than x and y at each unique point.

One option is to just precalculate density, which will work if z is indeed just a simple function of x and y:

ranges <- c(min(df$x, df$y), max(df$x, df$y))

n <- 1 + diff(ranges)/5

dens <- MASS::kde2d(df$x, df$y, n = n, lims = c(ranges, ranges))

cbind(expand.grid(x = dens$x, y = dens$y), 
      dens = c(dens$z), z = dens$x + dens$y) |>
  ggplot(aes(x, y)) +
  geom_tile(aes(fill = z, alpha = dens)) +
  theme_classic() +
  scale_alpha_continuous(range = c(0, 1)) +
  coord_equal()

If, in your actual data, z is an independent variable, then you may need to interpolate its value using something like interp::interp:

cbind(expand.grid(x = dens$x, y = dens$y), 
      dens = c(dens$z), 
      z = c(interp::interp(x = df$x, y = df$y, z = df$z, 
                           duplicate = "mean", xo = dens$x, yo = dens$y)$z)) |>
  ggplot(aes(x, y)) +
  geom_tile(aes(fill = z, alpha = dens)) +
  theme_classic() +
  scale_alpha_continuous(range = c(0, 1)) +
  coord_equal()

Which will give the same output in this particular instance.

本文标签： rHow to set alpha by bivariate density and fill by z in geomrasterStack Overflow