Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: One to N matching


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: One to N matching
Date   Tue, 18 Nov 2008 19:46:25 -0000

Thanks for the illumination (and elimination), and sorry for the red
herring. Matching joins survey on my personal list of "Even if you think
that exceptionally you think you understand the question, you are still
wrong". 

Nick 

Austin Nichols

Nick--
I don't think -duplicates- addresses the issue.  The idea is, for
observation 1, find the k obs with the closest value of p (k a scalar,
say 4, and p a variable, measuring for example a propensity).  Then
repeat for observation 2, but do not pick any of the obs you picked
for observation 1.  Et cetera. "Duplicates" in this setting means
picking observations more than once, I believe.

This is presumably needed for some nonstandard form of propensity
score matching--on which see Rosenbaum and Rubin (1983), linked
directly from p.26 of http://pped.org/stata/ciwod.pdf

On Tue, Nov 18, 2008 at 2:17 PM, Nick Cox <n.j.cox@durham.ac.uk> wrote:
> At least part of this sounds very easy. If duplicates are not allowed,
> remove them first with -duplicates-.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Gao Liu
>
> vmatch can match all observations based on certain criteria, like
> within some distance. However, what I want is the nearest N
> observations, where n is a fixed number. Also, duplicates are not
> allowed.

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



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index