How to extract outliers from box plot in R

Question

Could you explain me if there is a way to extract outliers from box plot. I have plotted a box plot and I want to extract only the outliers.

Here is the code for the box plot.

# melting down
require(reshape)
melt_nx <- melt(nx, id.vars = c("x", "y"))
boxplot(data = melt_nx, main = "NX", value ~ variable, las = 2,    
par(mar = c(15, 5, 4, 2) + 0.1),
names = c("We1", "We2", "we3"))

Is it possible from the box plot to extract the outliers only?

Answer 1

The boxplot function returns a list with one of it node-names as "out". These are the values that are beyond the "whiskers". I don't know about executing par within the argument list but if you want these particular values, then use this:

vals <- boxplot(data = melt_nx, main = "NX", value ~ variable, las = 2,    
      names = c("We1", "We2", "we3"))
vals$out

And do read all these help pages:

?boxplot
?boxplot.stats
?bxp
?fivenum

Answer 2

I know this has been answered, but for me there is an alternative method using the Boxplot method from the car package. Note the capital B in the Boxplot function call.

This is the code that does it for me, it returns the row numbers of the outliers which you can then use in your dataframe to filter out or extract, etc...

outliers<-Boxplot(x~y, data=df, id.method="y")

Note that the extracted values are of type Character. Then to exclude them you could do something like:

df2 <- df[-as.numeric(outliers),]

Hope this helps a little