[R] "continuous" boxplot?

[Deepayan Sarkar]

On 10/2/07, Bert Gunter <gunter.berton at gene.com> wrote:
>
>
> Folks:
>
> I found the references in the previous replies to this vexing data
> visualization issue to be quite interesting and useful. I think it fair to
> say that there is no single "best" way to do this -- it all depends on what
> you need to learn , and probably several alternative displays will be
> necessary to get the important information the data have to convey.
> However,as always, this issue has been considered before, and it may be
> worthwhile to at least consider an already available "standard" approach"
> using shingles and a trellis-type plot. ?xyplot and ?shingle should get you
> started (you probably want to shingle or bin on quantiles of y). The
> canonical reference is Bill Cleveland's VISUALIZING DATA (see "coplots").

A better reference for that is Elements of Graphing Data, where
boxplots created after discretizing the x variable is used to motivate
lowess. It's pretty much the only legitimate use I have seen of
boxplots with shingles (it's a cheap way to get both a sense of
location and variability non-parametrically). I vaguely remember
seeing an even better reference, but can't find it now.

-Deepayan