Nick's example just goes to show the issue isn't color vs. fill pattern,
it is careful thoughtful effort to communicate vs. thoughtless application
of whiz-bang "features," and sometimes "color" is one of those features
poorly applied. Thus, stata, wise in many ways, is not protecting anyone
by not providing fill patterns for graphs. Citing Tufte is informative,
and Tufte is smart, but I doubt he (or anyone) can specify what will
*always* be the absolute best way to display graphical information. Just
as the color choice Nick mentioned seems confusing and unwise, fill
patterns may have meaning in a community, and thus their deletion from
stata is problematic. For this and other reasons, in this case stata's
decision is not helpful, for it is just reducing the tools available for
the careful to thoughtfully make their effort to communicate. Not good.
Take care.
Sam
PS--For another maddening example, someone used blue for Democrats on maps
during the 2000 election in the US when Democrats are to the left of the
other dominant party and for centuries RED has been the color of the
parties of the left. Now, the conversation in the US is constantly
confusing, as some still say "better dead than red" while other right wing
supporters, avidly anti-communist, say "let's make 'America' red." Yes,
Tufte aside, it would have been far better if those maps had used fill
patterns rather than (the wrong) colors.
Take care.
Sam
On Wed, 6 Jun 2007, Nick Cox wrote:
> I didn't say StataCorp were right, although I usually think
> that. And I suppose it's good news that some keen Stata users seem
> to want to do everything in Stata. And if you used TeX, you
> could invent your own grammar, except that we would not
> let you publish it in the Stata Journal. Indeed, the people
> at StataCorp correct my grammar, and it ain't so bad.
>
> While I am talking graphics, I will share a recent discovery
> of what I consider dopey design. A recent introductory text has
> some colour code for distinguishing left- and right-skewed
> distributions. I am not sure I remember the details, because they
> were so arbitrary, but the flavour runs something like this:
>
> 1. Left-skewed distributions are shown in green.
>
> 2. Right-skewed distributions are shown in purple.
>
> So, histograms, distribution curves, and so forth, are coloured
> accordingly.
>
> This is extreme indirectness. Consider some poor student
> taking a test on Statistics 100. Here is a graph. It is in green.
> What does that mean? Is it left-skewed? Or right-skewed?
> (Or, more likely, green means something special, but I
> forget what....)
>
> Even if the whole world used this convention -- and clearly
> that is an enormous if -- it still sounds dopey. Why not
> just teach students how to tell the difference by looking
> at the distribution? The colour codes are just a diversion!
>
> I remain astonished at the well-known statistician who
> either invented this or was persuaded into it by his publisher.
>
> Not tarring anyone with this example, but it's too extraordinary
> not to share.
>
> Nick
> [email protected]
>
> Jeph Herrin
>
> > While I applaud the dedication of StataCorp to quality
> > presentation, many of us have to work with publications
> > that do not share this dedication. Just as I prefer Word
> > to allow me my grammatical errors, Stata would be more
> > valuable if it allowed me my errors of presentation. But
> > for top tier medical journals I must routinely resort to
> > Excel for those...
>
> Nick Cox wrote:
>
> > > Stata does not support stripes, whether straight, wavy, or of
> > > some other kind; nor does it support stipples, spots, polka dots,
> > > or other kinds of patterns. The good people at StataCorp read Edward
> > > Tufte's books, and in particular his admonitions about Moire
> > > vibration, and resolved not to do that.
> > >
> > > Please change your mail signature from "Stata List" as far
> > > as communications with this list are concerned. This
> > > may amuse you, or otherwise be congenial or convenient,
> > > but I suggest that it is an inappropriate identity for
> > > a member of Statalist. It is not compulsory, but it
> > > is an almost universal habit on this list to use real
> > > names and not hide identities.
> > >
> > > Nick
> > > [email protected]
> > >
> > > Stata List (a.k.a. AC)
> > >
> > >> I've been using Stata's graph capabilities to make amazing coloured
> > >> maps, with different colours for different indicators.
> > However, since
> > >> most printing is done on B&W printers, I would like to know how to
> > >> specify patterns (say, 45 degree stripes, or spots, etc.)
> > instead of
> > >> only solid colours.
> > >
> > > *
> > > * For searches and help try:
> > > * http://www.stata.com/support/faqs/res/findit.html
> > > * http://www.stata.com/support/statalist/faq
> > > * http://www.ats.ucla.edu/stat/stata/
> > >
> > >
> > *
> > * For searches and help try:
> > * http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
