geom_freqpoly shade quartiles

[Hkesh22]

Hi all,

I am fairly new to R. I am plotting a distribution using ggplot2, geom_freqpoly. I can add lines that show 0-25%, 25-50%, 50-75%, and 75-100% of the population, but I'd like to shade each quartile in different colours. I've attached a sample data set.

Here is my code:

library(ggplot2)

CD25 <- read.csv(file = "CD25 distribution Tregs.csv")

summary(CD25)


ggplot(CD25)+
  geom_freqpoly(aes(x=CD25.MFI, y =(..ncount..)),colour = "black",  binwidth = 0.01)+
  scale_x_log10()+geom_vline(xintercept = 52.718,col="grey",lwd = 0.5)+
  geom_vline(xintercept = 101.151,col="grey",lwd = 0.5)+
  geom_vline(xintercept = 165.561,col="grey",lwd = 0.5)+
  theme_bw()+theme(axis.ticks.length= unit(0.3,"cm"),panel.grid.major = element_blank (),panel.grid.minor = element_blank ())

Thanks for any help you can provide,

Hari

[eipi10]

For some reason, the site wouldn't allow me to place the final plot below the code, so the output is above. The code below uses the cut function to mark the quartiles. The interpolation is to ensure that there aren't any gaps in the color fill between quartiles.

p = ggplot(CD25)+
  geom_freqpoly(aes(x=CD25.MFI, y =(..ncount..)), colour = "black",  binwidth = 0.01)+

  scale_x_log10()+
  geom_vline(xintercept = 52.718,col="grey",lwd = 0.5)+
  geom_vline(xintercept = 101.151,col="grey",lwd = 0.5)+
  geom_vline(xintercept = 165.561,col="grey",lwd = 0.5)+
  theme_bw()+theme(axis.ticks.length= unit(0.3,"cm"),
                   panel.grid.major = element_blank (),
                   panel.grid.minor = element_blank ())

# Extract plot data
pb = ggplot_build(p)
d = pb$data[[1]][c("x","ncount","count")]

# Get quantile breaks
breaks = quantile(rep(d$x, d$count))

# Interpolate x and y values
xvals = seq(min(d$x), max(d$x),length=1000)
d = as.data.frame(approx(d$x, d$ncount, xout=xvals))

# Add grouping variable by quartile
d$quart = cut(d$x, c(-Inf, breaks[2:4], Inf))

# Fill plot by quartile
p + geom_area(data=d, aes(10^x, y, fill=quart), alpha=0.3, show.legend=FALSE)

[eipi10]

If you still want the vertical lines, you can do with a single call to geom_vline and you don't need to hard-code the xintercept values. For example p + geom_vline(xintercept=10^breaks[2:4], col="grey", lwd=0.5).