Create a ggplot2 stat_ecdf plot with standard error shading

Question

I have data from three doses of a treatment, with three replicates per each dose:

df <- data.frame(dose=c(rep(1,300),rep(3,300),rep(5,300)),
replicate=rep(c(rep("X1",100),rep("X2",100),rep("X3",100)),3),
value=c(rnorm(300,1,1),rnorm(300,3,1),rnorm(300,5,1)),stringsAsFactors=F)

df$dose <- factor(df$dose,levels=c(1,3,5))

I want to display it using an cdf plot. Per each replicate I can simply plot the cdfs of the three doses with:

for(r in c("X1","X2","X3")){
 ggplot(dplyr::filter(df,replicate==r),aes(x=value,color=dose))+
   stat_ecdf(geom="step")+ 
   theme_bw()+
   theme(panel.border=element_blank(),strip.background=element_blank())
}

But I'm looking for a way to display all replicates of each dose in one figure, with standard error shading around the mean value, similar to plots achieved with stat_smooth.

Can this be achieved?

Also, either for this or for a single replicate's plot:

r <- "X1"
ggplot(dplyr::filter(df,replicate==r),aes(x=value,color=dose))+
  stat_ecdf(geom="step")+ 
  theme_bw()+
  theme(panel.border=element_blank(),strip.background=element_blank())

Is there a way to compute the area under each of the ecdfs?

you can get the area under curve; read this [thread](https://stackoverflow.com/questions/4954507/calculate-the-area-under-a-curve) — M--, Jun 27 '17 at 17:37

M-- · Answer 1 · 2019-09-09T13:24:40.260

You can use interaction to group the data (in ggplot) based on two columns;

ggplot(df,aes(x=value,color = interaction(replicate, dose)
            , group=interaction(replicate, dose)))+
stat_ecdf(geom="step")+ 
theme_bw()+
theme(panel.border=element_blank(),strip.background=element_blank())

This would be your plot;

As it is a little bit vague:

if you want to have lines for dose then you can get rid of replicate in the interaction or use dose instead of replicate in your ddplyr::filter;

  ggplot(dplyr::filter(df,dose==1),aes(x=value,color = interaction(replicate, dose)
            , group=interaction(replicate, dose)))+
stat_ecdf(geom="step")+ 
theme_bw()+
theme(panel.border=element_blank(),strip.background=element_blank())

And you'd get:

Create a ggplot2 stat_ecdf plot with standard error shading

1 Answers1