The generic problem I think of here as "the way ahead is -bar-red",
meaning that -graph bar- or -histogram- doesn't do it, so what is
to be done? The answer is (usually) to make your own with -twoway bar-
as a basis.
Here is an example:
sysuse auto, clear
twoway__histogram_gen mpg if foreign, gen(freq_foreign x) freq start(0) width(2)
twoway__histogram_gen mpg if !foreign, gen(freq_domestic x2) freq start(0) width(2)
qui gen freq = freq_domestic + freq_foreign
twoway bar freq_domestic x if freq < ., barw(2) bcolor(blue) xsc(r(10 .)) || ///
rbar freq_domestic freq x2 if freq < ., barw(2) bcolor(red) ///
legend(order(1 "Domestic" 2 "Foreign") pos(11) col(1) ring(0)) ///
xtitle("`: var label mpg'") ytitle(Frequency) xla(10(5)40) yla(, ang(h)) ///
xsc(r(10 .))
The recipe here is
1. Use -twoway__histogram_gen- to generate the raw frequencies (densities,
percents). On this undocumented command, count the underscores carefully
(there are three, not two), see its help and also
SJ-5-2 gr0014 . . . . . . . Stata tip 20: Generating histogram bin variables
. . . . . . . . . . . . . . . . . . . . . . . . . . . . D. A. Harrison
Q2/05 SJ 5(2):280--281 (no commands)
tip illustrating the use of twoway__histogram_gen for
creation of complex histograms and other graphs or tables
itself reprinted in "33 Tips" (order one for everybody in your group as
a seasonal present).
2. We are going to stack frequencies, so we would need to generate every
layer in the stack but the first as the sum of all the previous.
3. We start with a -twoway bar- of the first category and then add others
using -twoway rbar-.
4. All else is cosmetics.
I'd be interested to hear Marcello's defence of this as compared with
alternatives. My prejudice is that it could be fairly successful with
a very small number of categories (two or three) but otherwise could
just look a mess.
Nick
[email protected]
> -tabplot- and -tableplot- from SSC will do
> something similar, arguably better.
>
> But in general the recipe is: work your
> way towards a -twoway bar-.
>
> I'll post some example code later.
Marcello Pagano
> > Has anyone got code for allowing a second, categorical
> > variable into the
> > histogram? What I am thinking of is the freq variable is
> the primary
> > variable going into the histogram, but there might be a
> > second variable,
> > age, say, categorised as `young' and `old'. Then think of
> each bar in
> > the histogram displayed as a two colour bar, black and white,
> > say, where
> > the black part represents the proportion of that bar who are
> > young, and
> > the white represents the proportion of that bar who are old.
> > A sort of
> > stacked-bar-meet-histogram kind of plot.
>
*
* 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/
