Bookmark and Share

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


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

Re: st: sampling using matching variables


From   Debs Majumdar <[email protected]>
To   [email protected]
Subject   Re: st: sampling using matching variables
Date   Tue, 17 Aug 2010 15:43:19 -0700 (PDT)

@Steve: I don't. But 2 of the reviewers wanted it as a part of sensitivity 
analyses. Thanks anyways.

-Debs




----- Original Message ----
From: Steve Samuels <[email protected]>
To: [email protected]
Sent: Tue, August 17, 2010 9:52:46 AM
Subject: Re: st: sampling using matching variables

Why do you require 1-1 matching?  There is more information (= greater
power, smaller standard errors) in analyses of matched sets with >1
person from each group.

If you insist, despite this advice, then let  "mset" be the variable
that identifies a matched set -will differ from command-to-command (I
like -cem- myself):

**********************************************
set seed 12345  // choose your own seed
gen u=uniform()
bysort mset race (u): keep if _n==1
*********************************************

Steve

-- 
Steven Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783



On Tue, Aug 17, 2010 at 12:29 PM, Christopher Swearingen
<[email protected]> wrote:
> Debs,
>
> I have used the contributed psmatch2, a propensity scoring method, to
> match groups on various covariates.
>
> findit psmatch2
>
> *** 1-1 age only matching
> psmatch2 nonwhite age, noreplacement
> gen age_matched = _weight
>
> *** 1-1 age & education matching
> psmatch2 nonwhite age education, noreplacement
> gen age_education_matched = _weight
>
> *** 1-1 age & education & gender matching
> psmatch2 nonwhite age education gender, noreplacement
> gen age_education_gender_matched = _weight
>
> ~c
> ________________________________________________________
> Christopher Swearingen, Ph.D.
> Assistant Professor of Pediatrics and Biostatistics
> University of Arkansas for Medical Sciences
> Department of Pedatrics
> One Children's Way, Slot 512-43
> Little Rock, AR 72202
>
> http://www.arpediatrics.org/research/biostatistics
>
>
>
> On Mon, Aug 16, 2010 at 10:27 PM, Debs Majumdar <[email protected]> wrote:
>> Hi,
>>
>>  I have a dataset of around 1000 people where 750 are whites and 250 are
>> non-whites. I have variables for age (35-75), education (0-20) and gender (0 
>or
>> 1). Also in the dataset I have item responses to 10 questions for each of 
>these
>> 1000 people.
>>
>>  I would like to create dataset(s) where one white is matched with a 
non-white
>> person in terms of age only, age and education, and age,education and gender.
>>
>> Initially, I tried matching on age only. But then for some cases I end up with 
>>3
>>
>> whites and 1 non-white for the same age category. At that time I would like 
to
>> drop any 2 of those 2 at random in the new dataset so that the matching is
>> always 1-1. How do I do that?
>>
>> I have tried using cem, vmatch etc. but didn't get what I want.
>>
>> Thanks,
>>
>> -Debs
>>
>>
>>
>> *
>> *   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/
>>
>
> *
> *   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/
>

*
*   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/



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