>
> > On the original question though, why isn't there something "off the
> shelf"
> > that will do what I want? Surely, a "boxplot" using mean, SD, max and
> min
> > would be a common enough need to justify it?
> >
>
> Gabor Grothendieck replied:
> tarr is not a list or a data frame. Use.data.frame(tarr) so that it
> uses the same assumptions as the examples in this thread.
>
> I believe there is no such facility due to a philosophical opposition.
> Unless someone were careful they would naturally assume that boxplots
> were shown even though that is not the case here.
>
> Let me elaborate on this a bit. In my opinion, this whole discussion is a
bad idea statistically/graphically (though of course on techniques for
implementation in R, it's fine). Boxplots were invented by John Tukey
SPECIFICALLY **not** to do as suggested above. They are a simple
implementation using quantiles to resistantly and graphically summarize
location, spread, shape (and can even give an indication of multi-modality)
of data. They also attempt to provide an indication of unusual values for
small data samples, but as with all things "outlier," this is rather
unreliable. Use of non-resistant summary statistics explicitly defeats these
purposes. Moreover, the evils of +/- 1 sd bars around the mean (which are
basically meaningless and generally mislead) have been noted for decades in
the statistical "culture" literature (i.e. the literature where
statisticians or others rail against bad statistical practice in science). I
think Bill Cleveland has ranted about this in times past, for example, but
my memory dims and I do not care to waste effort digging it up.
So I repeat: You have been shown by several how to make the plot. But don't.
Disclaimer: My view only.
Cheers,
Bert
>
Bert Gunter
Genentech Nonclinical Biostatistics
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
