Following an earlier reply to Neil Shephard, here
is a note on a further technique made possible
by Stata 9 for adding marginal information to graphs,
for example histograms.
This example concerns means, but the principles
can clearly be applied to other measures.
All the code here is given in one chunk at
the end for copying and pasting if desired.
Firing up the auto data,
. sysuse auto, clear
we need one variable that is identically zero
. gen zero = 0
and one that is identically some relatively
small negative number:
. gen minus = -0.25
"Relatively small" is judged with reference to
the range on the vertical axis: with a larger
maximum frequency than that to come, you would
need a much larger negative number; with
densities, usually a much smaller one.
Now we want the mean. In this first example, the
mean is naturally just a single value, and there
are other ways to do it, but in other examples
it is convenient to put results in a variable,
despite the redundancy implied. (That is also
why constants were put in variables above.)
. egen mean = mean(mpg)
Now we have all the ingredients:
. twoway histogram mpg,
w(1) freq legend(off) xti("Mileage (mpg)")
|| pcarrow minus mean zero mean,
barbsize(3) msize(3)
So the mean symbol is just an arrow that horizontally
is at the value of -mean- and vertically extends from
-minus- to -zero-. The arrow, however, is hidden by
its own rather large arrowhead, so a triangle alone is
visible.
You could -- especially in Stata 8 -- go for the
same effect by overplotting with a scatter plot with -ms(T)-,
but it can be troublesome to get the triangle in exactly
the right position and at exactly the right size.
Or at least that was my experience.
The same minor trickery can be used to populate
a series of histograms with their own mean symbols:
. egen meanby = mean(mpg), by(rep78)
. twoway histogram mpg, w(1) freq
by(rep78, legend(off)
subti("Mileage (mpg)", ring(1) pos(6)))
|| pcarrow minus meanby zero meanby,
barbsize(2) msize(2)
"To the vector belong the spoils."
(Norton Juster, "The dot and the line")
Nick
[email protected]
sysuse auto, clear
egen mean = mean(mpg)
gen zero = 0
gen minus = -0.25
twoway histogram mpg, w(1) freq legend(off) xti("Mileage (mpg)") ///
|| pcarrow minus mean zero mean, barbsize(3) msize(3)
more
egen meanby = mean(mpg), by(rep78)
twoway histogram mpg, w(1) freq by(rep78, legend(off) ///
subti("Mileage (mpg)", ring(1) pos(6))) ///
|| pcarrow minus meanby zero meanby, ///
barbsize(2) msize(2)
*
* 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/
