Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

re: st: randomly select n:1 matched controls per case without replacement?

From   "Ariel Linden. DrPH" <>
To   <>
Subject   re: st: randomly select n:1 matched controls per case without replacement?
Date   Sat, 26 Feb 2011 12:10:23 -0500

Hi Sandi,

It is not clear from your description if you are matching on an array of
covariates or just length of survival (which is not clear to me if you are
referring to age or survival from some treatment start date). If you have a
lot of matching variables, you should consider the propensity score and then
matching on that score. You can use PSMATCH2 (a user written program found
under -search all-) to choose 4 to 1 matches, but will require replacement
(I am not sure why the program was written that way, but I assume it is
because you can quickly run out of good matches and end up with too few
matched groups.

If you are only matching on length of survival, you can use cem (coarsened
exact matching - also a  user written program found at Gary King's website
at Harvard), or optmatch2 (another user written program - search optmatch2-)

I am not sure what you are trying to do here, since it seems you are
matching on the outcome and working backwards, as opposed to matching on
pre-intervention characteristics and looking forward based on a treatment
variable (as one would do in an RCT). That is, unless you are trying to
determine what were the characteristics that caused death differentially
between groups (ala case-control)? That is not clear from your posting.

I hope this helps


Date: Fri, 25 Feb 2011 07:15:35 -0600
From: "Pruitt, Sandi" <>
Subject: st: randomly select n:1 matched controls per case without

Im having problems figuring out how to select matching controls. Id like to
randomly select 4 living controls (dead=0) per each deceased case (dead=1).
I want to match, without replacement, on length of survival, measured as a
continuous measure of months (1, 1.1, 1.2etc.) (although I could recode into
days if that was easier). I want to try it 2 ways, matched exactly by month
(those living 1, 2, 3 months), and then matched by 6 month intervals (6, 12,
18.months). Any help is much appreciated.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index