Re: Question about boxplot

["Richard M. Heiberger" <rmh@temple.edu>]

tmp <- list(a=rnorm(20), b=rnorm(15))
boxplot(tmp)
boxplot(tmp, ylim=c(-1,1))


## Francisco J. Zagmutt's proposal to subset the data before sending
## it to bwplot won't work as that would change the locations of the
## quartiles.  To respond to his legitmate comment about potentially
## misleading the readers you will need to indicate truncation
## information, for example:

tmp.trunc <- do.call("cbind",
                     lapply(tmp,
                            ylim=c(-1,1),
                            function(x, ylim) {
                              c(high=sum(x > max(ylim)),
                                  low=sum(x < min(ylim)))
                            })
                     )
tmp.trunc

tmp.x <- boxplot(tmp, ylim=c(-1,1))
axis(side=1, at=tmp.x, tmp.trunc["low",], pos=-.9, ticks=FALSE)
axis(side=3, at=tmp.x, tmp.trunc["high",], pos=.9, ticks=FALSE)


## Rich