R – Shade (fill or color) area under density curve by quantile

ggplot2r

Suppose e.g. I want to shade the area under the density curve for the standard normal distribution by decile. I want the left-most 10% of the area to have a different shading to the next 10% and so on.

This is a variant on the questions "Shading a kernel density plot between two points" and "ggplot2 shade area under density curve by group", but I want to shade each quantile (in my example, each group is a decile but the process should easily generalise to other quantiles).

I don't mind whether a solution uses ggplot2 or base graphics, and whether this is done directly from a formula (which would be really neat) or based on making a data frame first. If the latter, you may want:

delta <- 0.0001 
z.df <- data.frame(x = seq(from=-3, to=3, by=delta))
z.df$pdf <- dnorm(z.df$x)
z.df$decile <- floor(10*pnorm(z.df$x) + 1)

Note that the naive solution ggplot(z.df, aes(x = x, fill = quantile)) + geom_ribbon(aes(ymin = 0, ymax = pdf)) would fail because Aesthetics can not vary with a ribbon.

Best Answer

Actually aesthetics can vary with geom_ribbon(...) (or geom_area(...), which is basically the same thing), as long as you set the group aesthetic as well.

delta     <- 0.001 
quantiles <- 10
z.df     <- data.frame(x = seq(from=-3, to=3, by=delta))
z.df$pdf <- dnorm(z.df$x)
z.df$qt  <- cut(pnorm(z.df$x),breaks=quantiles,labels=F)

library(ggplot2)
ggplot(z.df,aes(x=x,y=pdf))+
  geom_area(aes(x=x,y=pdf,group=qt,fill=qt),color="black")+
  scale_fill_gradient2(midpoint=median(unique(z.df$qt)), guide="none") +
  theme_bw()

Setting quantiles <- 20 at the beginning produces this:

Related Solutions

R – ggplot2 shade area under density curve by group

Here is one way (and, as @joran says, this is an extension of the response here):

#  same data, just renaming columns for clarity later on
#  also, use data tables
library(data.table)
set.seed(1)
value <- c(rnorm(50, mean = 1), rnorm(50, mean = 3))
site  <- c(rep("site1", 50), rep("site2", 50))
dt    <- data.table(site,value)
#  generate kdf
gg <- dt[,list(x=density(value)$x, y=density(value)$y),by="site"]
#  calculate quantiles
q1 <- quantile(dt[site=="site1",value],0.01)
q2 <- quantile(dt[site=="site2",value],0.75)
# generate the plot
ggplot(dt) + stat_density(aes(x=value,color=site),geom="line",position="dodge")+
  geom_ribbon(data=subset(gg,site=="site1" & x>q1),
              aes(x=x,ymax=y),ymin=0,fill="red", alpha=0.5)+
  geom_ribbon(data=subset(gg,site=="site2" & x<q2),
              aes(x=x,ymax=y),ymin=0,fill="blue", alpha=0.5)

Produces this:

R – ggplot2: fill color behaviour of geom_ribbon

An option that doesn't require manually creating another column would be to do the logic within aes(fill = itself;

## fill dependent on level > level2
h + 
  geom_ribbon(aes(ymin = level, ymax = level2, fill = level > level2)) +
  geom_line(aes(y = level)) + geom_line(aes(y=level2)) +
  scale_fill_manual(values=c("red", "green"), name="fill")

Or, if you only want to fill based on that condition being true,

## fill dependent on level > level2, no fill otherwise
h + 
  geom_ribbon(aes(ymin = level, ymax = level2, fill = ifelse(level > level2, TRUE, NA))) +
  geom_line(aes(y = level)) + geom_line(aes(y=level2)) +
  scale_fill_manual(values=c("green"), name="fill")

I assume the lack of interpolated fill seems to have something to do with the ggplot2 version, as I get the same thing happening with @beetroot's code

## @beetroot's answer
huron$id <- 1:nrow(huron)
huron$group <- ifelse(huron$id <= 50, "A", "B") 

h <- ggplot(huron, aes(year))
h +
  geom_ribbon(aes(ymin = level, ymax = level2, fill = group)) +
  geom_line(aes(y = level)) + geom_line(aes(y = level2))

I get @ManuK's image output when running that code without logic in aes(fill =.

Best Answer

Related Solutions

R – ggplot2 shade area under density curve by group

R – ggplot2: fill color behaviour of geom_ribbon

Related Topic