Re: st: * mark indicating between group significans in box plots

[Ronan Conroy <rconroy@rcsi.ie>]

On 2011 Samh 10, at 11:50, Nick Cox wrote:

> I've often seen people test a difference between means and show a box
> plot as the graphical equivalent. I'd argue that a plot of means with
> confidence intervals would be directly related, although even then
> there are pitfalls.
> 
> Of  course, that may not be what you are doing. If your test were say
> a Mann-Whitney-Wilcoxon test, a box plot makes more sense.

There was a vogue for a notched boxplot that allowed a visual comparison of medians. The 'notch' is the region

median ± 1.58 x (H-spread)/sqrt(n)

Two groups whose notched intervals do not overlap have significantly different medians, though this procedure does not adjust for multiple comparisons. 

McGill R, Tukey JW, Larsen WA. Variations of box plots. The American Statistician 1978 1978;32:12-16

I haven't seen one in years, though.

Ronán Conroy
rconroy@rcsi.ie
Associate Professor
Division of Population Health Sciences
Royal College of Surgeons in Ireland
Beaux Lane House
Dublin 2


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/