RE: st: Combining a survey weight and a frequency weight

[Joe Canner <jcanner1@jhmi.edu>]

Perhaps what James is referring to is that -psmatch2- (in the case of 1:many matching) gives non-integer weights (each treated case gets a weight of 1 and each control gets a weight of the reciprocal of the number of control matches).  These, of course, cannot be used with [fweight=], but can be used with importance weights [iweight=].  I am currently doing an analysis using -psmatch2- followed by Cox regression and it appears that [iweight=] and [pweight=] give the same results.  The choice and implication of weighting method may differ for other procedures.

Regards,
Joe Canner

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Radwin, David
Sent: Monday, August 26, 2013 12:59 PM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: Combining a survey weight and a frequency weight

Please note that the FAQ asks you to mention authors and locations of user-written programs (-psmatch2- is by Edwin Leuven and Barbara Sianesi and is available from SSC).

Although the weight variable in -psmatch2- does represent a frequency, it is not a frequency weight in the sense of Stata's -fweight-. I would not describe it as representing "how important the observation is." 

Here is the description in the program's help file:


   psmatch2 creates a number of variables for the convenience of the user . . . . 
        _weight. For nearest neighbor matching, it holds the frequency with which the observation is
        used as a match; with option ties and k-nearest neighbors matching it holds the normalized
        weight; for kernel matching, and llr matching with a weight other than stata's tricube, it
        stores the overall weight given to the matched observation. When estimating att only _weight =
        1 for the treated.


In other words, I don't think you should try to use this variable as a measure of importance.

David
--
David Radwin
Senior Research Associate
Education Studies Division
RTI International
2150 Shattuck Ave., Suite 800
Berkeley, CA 94704
Phone: 510-665-8274

www.rti.org


> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner- 
> statalist@hsphsun2.harvard.edu] On Behalf Of James Jensen
> Sent: Saturday, August 24, 2013 6:49 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Combining a survey weight and a frequency weight
> 
> I would like to add that the frequency weight is generated by 
> propensity score matching by the psmatch2 command. I found out earlier 
> that this weight is a frequency weight that is to tell how important 
> Stata how important the matched observation is.
> 
> I would like to combine this frequency weight and the survey weight 
> when I carry out the analysis on the matched sub-sample.
> 
> But I am not sure what the weight type should be.
> 
> 
> ----- Original Message -----
> From: James Jensen <jamesjensen212@yahoo.com>
> To: "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
> Cc:
> Sent: Saturday, 24 August 2013, 14:35
> Subject: st: Combining a survey weight and a frequency weight
> 
> Hi all,
> 
> I have a survey weight that accounts for the non-response rates of the 
> given observations, and the weights sum to the population total.
> 
> I have also a frequency weight that is given in the Dataset to tell 
> Stata how important the observation is.
> 
> If I want to combine these two weights into one single weight by 
> multiplying them together, what will the weight type become? Is it 
> going to be a frequency weight or a probability weight.

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