Re: st: Boxplot - color coded + cat labels | documenting work round

["Allan Reese (Cefas)" <allan.reese@cefas.co.uk>]

Thank you Sergiy - you spotted what I spotted immediately after sending the message - that I got the pink and blue wrong way round.  I corrected it for my customer and decided against sending my mea culpa to everyone. "Be sure your sins will find you out."  

As you also correctly point out, the repeated MF labels might be considered fussy, but to me is justified to avoid either constantly referring to a legend or having to memorize the legend.  In my actual graph at the higher ages there is only one sex (in the sample).  Hence adjusting the gap() improves understanding by distinguishing MF pairs at each age but breaks down at the end because of nofill, which is needed because of the yvars/over combination.  I'm an advocate of intended redundancy for reinforcement in graphs (especially as color may be printed in B&W). Tufte is more grudging:

"Multiple signals will help escape from the swamp of perceptual shifts and other ambiguities in reading. Redundant and partially overlapping methods of data presentation can yield a sturdy design, responding in some way of another to potential visual complications - with, however, a resulting danger of fussy, cluttered, insecure, committee-style design.  A crystalline, lucid redundancy will do." (Envisioning Information, p93)

I have also asked StataCorp for more plotting symbols.  You can get mars and venus as a free font and use them as centred labels as a work-round.  I will send that workround in a separate list message so it's indexed.

Allan



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