st: Re: Problems with nnmatch

["Michael Blasnik" <michael.blasnik@verizon.net>]

I also just tried using nnmatch this past week and found it too slow -- I 
stopped it before ever getting results.  I took a look at the code and 
realized that it would be a big project to make it faster, but I think that 
substantial speed improvements (maybe 10x) could be made in current ado code 
and perhaps more using Mata.  The analysis approach pretty much requires 
looping across each observation to locate matches, so it will still be time 
consuming, but hopefully measured in minutes instead of hours.  I may try to 
make some speed improvements on my own some time in the next couple of 
months.

Michael Blasnik
michael.blasnik@verizon.net


----- Original Message ----- 
From: "Carlos Osorio" <cosoriou@MIT.EDU>
To: <statalist@hsphsun2.harvard.edu>
Sent: Tuesday, July 26, 2005 2:49 PM
Subject: st: Problems with nnmatch



> Hi all,
>
> I'm using the nnmatch to analyze average treatment effects, but have not 
> been able to get a result after running it for 5+ hours.  I'm using a 
> sample with about 30,000 observations, and set memory to 1400 Megs, 
> without luck. I've made a smaller subsample (random) and no luck either.
>
> Does any one have had an experience like this? If so, can you give some 
> hints on how to solve it?
>
> Many thanks!
>
> - Carlos
>
>
> Carlos A. Osorio
> ===================
> Ph.D. Candidate
> MIT Engineering Systems Division
> http://esd.mit.edu/
*
*   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/