GGplot with multiple/conditional scale_fill_manual for stat_density_ridges - r

Question

Here's a question for ggplot experts...
My dataset has 432000 observations of 4 variables (one is numeric, the others are factors). Predictors has 6 levels, Estimate has 4 levels, Model has 2 levels. Value has a max of 2.6 and a min of -3. (I hope you can create data with that information.)
The plot set-up is a 4x6 faceted plot here's is a 2x3 example of the plot:
each row is a different level of a factor (Predictors)
each column a different level of another factor (Estimate)
there are two distributions within each mini-plot which is another factor (Model)
The goal is to plot:
the distributions in each column in a different color (blue, green, red, yellow) (according to Estimate)
within each mini-plot, the shade/hue of that color should be different (e.g., within the green column, repeat the order of colors according to Model)
fill the tails of two quantiles on each distribution of each mini-plot (as the tail lines in the picture indicate; color the tail from each line to the end of the tail in black/gray). The tails can be the same throughout the entire plot.
Here's an example of the code that I'm using. It doesn't plot the quantiles in a separate color:
pp <- ggplot(dd, aes(x=Value, y=as.factor(Model), fill=factor(Model))) +
stat_density_ridges(quantile_lines = TRUE, quantiles = c(0.05, 0.95), alpha = 0.95,vline_size = 0.5)+
scale_fill_manual(values = c("red", "white")) +
geom_vline(xintercept = 0, linetype="dashed", color = "black", size=0.5) +
facet_grid(Predictors~Estimate, scales = "free") + labs(x="Parameter value", y=" ") +
theme(text = element_text(size = 16)) + theme(axis.title=element_text(face="bold"), strip.text = element_text(
size = 16)) + theme(legend.position = "none")
To color the quantiles, you can swap fill=factor(Model) with fill=factor(..quantile..), but getting both "fills" in the same plot has been impossible thus far. Among many other things, I tried entering multiple factors into "fill", like this: fill=c(factor(Model), factor(Estimate), ..quantile..) , but it didn't work.
Any ideas?

Answer 1

I think from your description your data looks a bit like this (though I've limited it to 6000 rows):
set.seed(69)
Value <- rnorm(6000)
Predictors <- factor(rep(LETTERS[1:6], each = 1000))
Estimate <- factor(rep(rep(letters[1:4], each = 250), 6))
Model <- factor(rep(rep(c("Model1", "Model2"), each = 125), 24))
Value <- Value + rep(rnorm(6), each = 1000)
Value <- Value + rep(rep(rnorm(4), each = 250), 6)
Value <- Value + rep(rep(rnorm(2), each = 125), 24)
dd <- data.frame(Value, Predictors, Estimate, Model)
It sounds like most of what you want to do can be achieved by creating a new factor variable that is a conjunction of two other factors:
dd$fill_factor <- as.factor(paste0(Model, Estimate))
Which means that we should get close to the desired effect with minimal changes to your code:
library(ggplot2)
library(ggridges)
my_colors <- c("#0000FF", "#00FF00", "#FF0000", "#FFFF00")
ggplot(dd, aes(x = Value, y = Model, fill = fill_factor)) +
stat_density_ridges(quantile_lines = TRUE,
quantiles = c(0.05, 0.95),
alpha = 0.95,
vline_size = 0.5) +
scale_fill_manual(values = c(gsub("0", "6", my_colors),
gsub("F", "A", my_colors))) +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", size = 0.5) +
facet_grid(Predictors ~ Estimate, scales = "free") +
labs(x = "Parameter value", y = " ") +
theme(text = element_text(size = 16),
axis.title = element_text(face = "bold"),
strip.text = element_text(size = 16),
legend.position = "none")