Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: overlapping histograms


From   Jeph Herrin <junk@spandrel.net>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: overlapping histograms
Date   Tue, 30 Sep 2008 14:30:03 -0400

This audience knows about normal curves, and histograms,
and tails. That's about all they know, so I think the
graph I want is the right one. Anyway, it's not my
presentation.

I was deliberately ambiguous about "the nation" in the
interests of data privacy. If you want to guess I meant
the USA, then I suppose that reflects your bias, not
mine.... :)

P.S. I mentioned stripes to amuse, not to proselytize,
as we've been over that ground often enough before.


Nick Cox wrote:
Audience is crucial -- no disagreement there.

But your very clear implication is that you want to look at shifts in distribution and see how far they are consistent across the range: an excellent substantive question.

I know no more direct way to do that than to look at quantile functions, or equivalently their inverses. Histograms just obscure the details through arbitrary decisions on bin width and origin.

References to "the nation" are ambiguous on an international list. I guess in your case you are discussing the USA.

P.S. No votes from me for stripes, polka dots or other such graphjunk.

Jeph Herrin wrote:

Nick Cox wrote:
 > It is good to know that you can do this. My guess is that many users
 > would like to see it written up somewhere.
 >
 > However, these histograms tend to lead only to crude inferences. Users
 > can't read off quantitative conclusions from the graph, just stuff
 > like "The distributions are about the same", or "... very different",
 > etc.


It all depends on the audience, doesn't it? This is for a
presentation about quality improvement in health care for
a certain treatment. The mantra for many years has been
that you don't want to fix just the poor performers - you
want to "shift the curve". It's the right mantra, because
every patient benefits (here, from faster treatment times);
and for the mantra, this is the right figure to show people
responsible for improving patient care. In three years, the
national curve shifted noticeably.

Not exactly on list topic, but the point is that Stata is
only improved by letting users do what they want in graphics
(striped bars comes to mind...).
*
*   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/

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index