full text label on Boxplot, with added mean point

问题:

Am trying to get text label similar to what this https://stats.stackexchange.com/questions/8206/labeling-boxplots-in-r, but I cant get it to work. MWE similar to what I have is this:

data <- data.frame(replicate(5,sample(0:100,100,rep=TRUE)))

meanFunction <- function(x){
return(data.frame(y=round(mean(x),2),label=round(mean(x,na.rm=T),2)))}

ggplot(melt(data), aes(x=variable, y=value)) +
geom_boxplot(aes(fill=variable), width = 0.7) +
stat_summary(fun.y = mean, geom="point",colour="darkred", size=4) +
stat_summary(fun.data = meanFunction, geom="text", size = 4, vjust=1.3) 

That produces something like "A" in the attached image, and I am trying to get something like "B" for each of the boxes. Thanks.

回答1:

Here is my attempt. First, I reshaped your data. Then, I produced your boxplot. I changed the size and colour of text for mean. Then, I looked into the data that ggplot used, which you can access using ggplot_build(objectname)$data[[1]]. You can see the numbers you need. I selected necessary variables and reshaped the data, which is df. Using df, you can annotate the numbers you want.

library(dplyr)
library(tidyr)
library(ggplot2)

set.seed(123)
mydf <- data.frame(replicate(5,sample(0:100,100,rep=TRUE)))

mydf <- gather(mydf, variable, value)

meanFunction <- function(x){
return(data.frame(y=round(mean(x),2),label=round(mean(x,na.rm=T),2)))}

g <- ggplot(data = mydf, aes(x = variable, y = value, fill = variable)) +
     geom_boxplot(width = 0.5) +
     stat_summary(fun.y = mean, geom = "point",colour = "darkred", size=4) +
     stat_summary(fun.data = meanFunction, geom ="text", color = "white", size = 3, vjust = 1.3)

df <- ggplot_build(g)$data[[1]] %>%
       select(ymin:ymax, x) %>%
       gather(type, value, - x) %>%
       arrange(x)

g + annotate("text", x = df$x + 0.4, y = df$value, label = df$value, size = 3)

回答2:

First, I would take your data and then calculate all the boxplot features yourself. Here's one way to do that

dd <- data.frame(replicate(5,sample(0:100,100,rep=TRUE)))

tt <- data.frame(t(sapply(dd, function(x) c(boxplot.stats(x)$stats, mean(x)))))
names(tt) <- c("ymin","lower","middle","upper","ymax", "mean")
tt$var <- factor(rownames(tt))

I'm sure there are prettier ways to do that with dplyr but this point is you'll need to calculate those values yourself so you know where to draw the labels. Then you can do

ggplot(tt) + 
   geom_boxplot(aes(x=var, ymin=ymin, lower=lower, middle=middle, upper=upper, ymax=ymax), stat="identity", width=.5) + 
   geom_text(aes(x=as.numeric(var)+.3, y=middle, label=formatC(middle,1, format="f")), hjust=0) +
   geom_text(aes(x=as.numeric(var)+.3, y= lower, label=formatC(lower,1, format="f")), hjust=0) +
   geom_text(aes(x=as.numeric(var)+.3, y= upper, label=formatC(upper,1, format="f")), hjust=0) + 
   geom_text(aes(x=as.numeric(var)+.3, y= ymax, label=formatC(ymax,1, format="f")), hjust=0) + 
   geom_text(aes(x=as.numeric(var)+.3, y= ymin, label=formatC(ymin,1, format="f")), hjust=0) + 
   geom_point(aes(x=var, y=mean)) + 
   geom_text(aes(x=as.numeric(var), y= mean, label=formatC(mean,1, format="f")), hjust=.5, vjust=1.5)

to draw each of the labels