Neither this post nor this post apply to my case.
Assume:
set.seed(42)
x<-rep(c("A","B","C"), c(3,4,1))
y<-rep(c("V","W"),c(5,3))
z<-rnorm(8,-2,1)
df<-data.frame(x,y,z)
boxplot(z~x+y,df)
I want my plot to include groups with more than, say, one element. This means that I want my plot show only A.V, B.V and B.W. Furthermore, since my graph has about 70 groups, I don't want to do it by writing a list by hand.
Thanks
You can create a new column ('xy') using paste, create a logical index using ave for 'xy' groups having more than one elements, and then do the boxplot.
df1$xy <- factor(paste(df1$x, df1$y, sep='.'))
index <- with(df1, ave(1:nrow(df1), xy, FUN=length))>1
boxplot(z~xy, droplevels(df1[index,]))
Or using ggplot
library(dplyr)
library(tidyr)
library(ggplot2)
df %>%
group_by(x,y) %>%
filter(n()>1) %>%
unite(xy, x,y) %>%
ggplot(., aes(xy, z))+
geom_boxplot()
You can see if any bp$n are 0 and subset by that
set.seed(42)
df <- data.frame(x = rep(c("A","B","C"), c(3,4,1)),
y = rep(c("V","W"),c(5,3)),
z = rnorm(8,-2,1))
bp <- boxplot(z ~ x + y, df, plot = FALSE)
wh <- which(bp$n == 0)
bp[] <- lapply(bp, function(x) if (length(x)) {
## `bp` contains a list of boxplot statistics, some vectors and
## some matrices which need to be indexed accordingly
if (!is.null(nrow(x))) x[, -wh] else x[-wh]
## some of `bp` will not be present depending on how you called
## `boxplot`, so if that is the case, you need to leave them alone
## to keep the original structure so `bxp` can deal with it
} else x)
## call `bxp` on the subset of `bp`
bxp(bp)
Or you can use any value you like:
wh <- which(bp$n <= 1)
bp[] <- lapply(bp, function(x) if (length(x)) {
if (!is.null(nrow(x))) x[, -wh] else x[-wh]
} else x)
bxp(bp)
