st: Graphing RAREA with two crossing functions

[Ernest Berkhout <ernestb@seo.fee.uva.nl>]

The idea seems simple, the implementation that I tried so far appears to be 
not... Therefor I firstly report what I have managed to tackle myself, but 
finally there are still some open ends. Anyone out there who might have a clue?

The problem is as follows: imagine a twoway (overlayed) graph with two 
lines that cross eachother (maybe more than once), for example Y = sin(x) 
and Y2 = 0.2.
Now I want to shade the closed areas between these two lines. If possible, 
I also want different colors for the two possibilities Y>Y2 and Y<Y2. My 
first approach was to copy minimum and maximum values and then use the 
-twoway rarea- approach.

***********************************
The example can be reproduced with:

set obs 100
gen x = _n/10
gen y = sin(x)
gen y2=.2

*** create low/high-series for the areas where Y>Y2:
gen low1  = y2 if y > y2
gen high1 = y  if y > y2

*** create low/high-series for the areas where Y<Y2:
gen low2  = y  if y < y2
gen high2 = y2 if y < y2

twoway (line y x) (line y2 x ) (rarea low1 high1 x ) (rarea low2 high2 x )
********************************

This yields two problems:
1.      The shaded area of this graph does not extend until the exact 
points of intersection, especially when there are relatively few 
datapoints. The 'edges' of the area are left blank.
2. Follows from problem 1: although the variables low1 and high1 are set to 
missing when the values of Y are below Y2 (low1>high1), there is still an 
area to plot for Stata, namely from the last non-missing observations to 
the first non-missing observations from the next area where Y>Y2 again.

I managed to work around most of the second problem, by setting low1 & 
high1 equal to either Y or Y2 instead of missing. Then the code reads:

gen low1  = y2 if y > y2
gen high1 = y  if y > y2
replace low1 = y2 if low1==.
replace high1= y2 if high1==.

But when datapoints aren't many (or when functions aren't that smooth as in 
my example), problem 1 still leaves behind its marks. What is really needed 
is a way to adress the graph to fill the area up untill the intersection 
points (and not any further).
Does anyone have a clue? Is it possible to adress the intersection in a 
continuous way, or does one manually has to compute these points and add 
them (temporarily) to the data? Any opinions are welcome!



*
*   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/