st: Speeding up tw contour

["Dimitriy V. Masterov" <dvmaster@gmail.com>]

I am trying to plot the average binary outcome for two continuous
variables for 8,500 observations.

The data looks like this:

        z         y               x
        0   .4557176   97.56097
        1   2.096701   92.85714
        1   3.539873   92.72997
        1   1.989236   96.77419
        0   .0043634    86.8223
        0   .0009028   76.92308
        0   .0082523   71.42857
        0   .0262847   92.52631
        0   .7788773         25
        0   .3779977   96.15385

The range of x is (0,100] and y is (0,10]. The (x,y) grid is not
regular, so some interpolation is needed. In particular, data near the
top of the ranges is pretty sparse.

I tried -tw contour z y x-, but it has been running for over three
days on a very fast machine. I also tried changing the interpolation
method to Shepard, with similar results. I tried limiting the sample,
and nothing above 250 observations has yet to finish.

Is there any preprocessing I could do to my data to speed things up?

DVM
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/