Yoshiro-
The -locpoly- command runs in C rather than in interpreted code, so it
is much much faster than -lowess-, and you can also specify more
options (e.g. which kernel to use). You might try using
round(population/100) as the weight var you supply to -expand- to
speed up your intermediate work... if you think that small differences
in the population size used as weights are unlikely to change
parameter estimates much (but note that fweights that are zero are not
dropped by the -expand- step, so you should drop them in another step,
if that does not give you pause, or divide by a more appropriate power
of ten before rounding to the nearest integer).
On 5/2/06, Yoshiro Nagao <dengue_aedes@yahoo.co.jp> wrote:

Dear Nick, Austin, and Marrten,
Thanks again for your quick advice.
As Nick warned of, "expand" generate
580 000 observations, and a single
execution of "lowess" took 10 minutes.
But I am happy with it as far as it works.
What is the advantage of locpoly
over lowess?

*
* 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/