Feature request: Scaled densities/counts in 2d density/bins plots.

[bjreisman]

2d density plots are one of the most common data-visualizations used to display flow cytometry data, and the geom_bin2d and geom_hex and geom_density_2d geoms are excellent for making these plots. However, when facetting 2d density plots, there isn't a straightforward way to set the scale such that the highest point of each plot is the same - the convention in my field.

geom_density (the 1d version) has an option to set fill = ..scaled.. which rescales all groups to a maximum of 1. Would you consider adding such an option to geom_bin2d and geom_hex (and stat_bin2d/stat_binhex)?

library(ggplot2)
library(dplyr)
library(viridis)

#2d density plot colored by level
plot.density2d_level <- diamonds %>%
  ggplot(aes(x=x, y= depth)) + 
  stat_density_2d(aes(fill = stat(level)),
                  geom = "polygon", 
                  n = 100 ,
                  bins = 10) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")

plot.density2d_level

As you can see, its hard to tell where the area of highest density is in all but VVS1 and IF.

However, if I render the first panel alone, the area of highest density is very clear.

diamonds %>%
  filter(clarity == "I1") %>%
  ggplot(aes(x=x, y= depth)) + 
  stat_density_2d(aes(fill = stat(level)),
                  geom = "polygon", 
                  n = 100 ,
                  bins = 10) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")

If I set the fill to stat(piece) then I'm closer to achieving the right scaling; though there are still some panels which are not using the full color palette. (This has been my temporary solution.)

plot.denisty2d_piece <- diamonds %>%
  ggplot(aes(x=x, y= depth)) + 
  stat_density_2d(aes(fill = stat(piece)),
                  geom = "polygon", 
                  n = 100, 
                  bins = 10, 
                  contour = T) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")
plot.denisty2d_piece

stat_bin2d and stat_hex also lack the ability to set the fill to a 'scaled density' such as 'ndensity' or 'ncount'

plot.bin2d <- diamonds %>%
  ggplot(aes(x=x, y= depth)) + 
  stat_bin2d(bins = 60, aes(fill = stat(count))) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")

str(ggplot_build(plot.bin2d)$data[[1]])
#> 'data.frame':    2716 obs. of  20 variables:
#>  $ fill    : chr  "#000004" "#000004" "#000004" "#000004" ...
#>  $ xbin    : int  27 49 38 48 54 34 42 43 48 53 ...
#>  $ ybin    : int  22 22 23 23 23 24 24 24 24 24 ...
#>  $ value   : num  1 1 1 1 1 2 1 1 1 1 ...
#>  $ x       : num  4.74 8.68 6.71 8.5 9.58 ...
#>  $ y       : num  55.5 55.5 56.1 56.1 56.1 ...
#>  $ count   : num  1 1 1 1 1 2 1 1 1 1 ...
#>  $ density : num  0.00135 0.00135 0.00135 0.00135 0.00135 ...
#>  $ PANEL   : Factor w/ 8 levels "1","2","3","4",..: 1 1 1 1 1 1 1 1 1 1 ...
#>  $ group   : int  -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 ...
#>  $ xmin    : num  4.65 8.59 6.62 8.41 9.49 ...
#>  $ xmax    : num  4.83 8.77 6.8 8.59 9.67 ...
#>  $ ymin    : num  55.2 55.2 55.8 55.8 55.8 ...
#>  $ ymax    : num  55.8 55.8 56.4 56.4 56.4 ...
#>  $ colour  : logi  NA NA NA NA NA NA ...
#>  $ size    : num  0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 ...
#>  $ linetype: num  1 1 1 1 1 1 1 1 1 1 ...
#>  $ alpha   : logi  NA NA NA NA NA NA ...
#>  $ width   : logi  NA NA NA NA NA NA ...
#>  $ height  : logi  NA NA NA NA NA NA ...

plot.hex <- diamonds %>%
  ggplot(aes(x=x, y= depth)) + 
  stat_binhex(bins = 60, aes(fill = stat(count))) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")

str(ggplot_build(plot.hex)$data[[1]])
#> 'data.frame':    3005 obs. of  10 variables:
#>  $ fill   : chr  "#000004" "#000004" "#000004" "#000004" ...
#>  $ x      : num  4.74 8.5 8.68 6.62 9.49 ...
#>  $ y      : num  55.7 55.7 55.7 56.2 56.2 ...
#>  $ density: num  0.00135 0.00135 0.00135 0.00135 0.00135 ...
#>  $ count  : int  1 1 1 1 1 2 1 1 1 1 ...
#>  $ PANEL  : Factor w/ 8 levels "1","2","3","4",..: 1 1 1 1 1 1 1 1 1 1 ...
#>  $ group  : int  -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 ...
#>  $ colour : logi  NA NA NA NA NA NA ...
#>  $ size   : num  0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ...
#>  $ alpha  : logi  NA NA NA NA NA NA ...

Compare this to the 1d stat_density and stat_histogram which both provide the option to rescale each facet to a maxium of 1 using the y = stat(scaled) and y = stat(ndensity) option respectively.

plot.density <- diamonds %>%
  ggplot(aes(x=x)) + 
  geom_density(aes(y = stat(scaled))) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")
plot.density

plot.histogram <- diamonds %>%
  ggplot(aes(x=x)) + 
  geom_histogram(bins = 60, aes(y = stat(ndensity))) + 
  facet_wrap(clarity~.) + 
  scale_fill_viridis(option = "A")

plot.histogram

Would you consider adding a "scaled" statistic to stat_density2d and an "ndensity/ncount" statistic to stat_bin2d and stat_binhex?

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/