Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Districts in PSM
Austin Nichols <firstname.lastname@example.org>
Re: st: Districts in PSM
Fri, 30 Sep 2011 10:38:08 -0400
Henrik Wiig <email@example.com>:
On Thu, Sep 29, 2011 at 5:05 PM, Henrik Wiig <firstname.lastname@example.org> wrote:
> Hi Austin!
> Thanks for your response to my idea of PSM at district level and then aggregate to a total effect. As a not too experienced Stata user I'm afraid I need to find some more detailed description of how to go about doing this. However, searching statalist archieve using different combinations of the term givens in your response, I am not able to find the examples you are refering too. Could you help out a "blind" man by giving more details to the procedure to follow?
> Fra: email@example.com [firstname.lastname@example.org] på vegne av Austin Nichols [email@example.com]
> Sendt: 27. september 2011 17:36
> Til: firstname.lastname@example.org
> Emne: Re: st: Districts in PSM
> Henrik Wiig <email@example.com> :
> Several suggestions have appeared on Statalist over the last few
> years, including computing a pseudo-score that is the propensity score
> plus 100 times an identifier, ensuring that matches are made (e.g. by
> -psmatch2- findable via findit) within distinct values of the
> identifier. I prefer reweighting, which can also be done within
> district, though one would have to be clear about what the estimand
> is, up front.
> On Tue, Sep 27, 2011 at 11:27 AM, Nick Cox <firstname.lastname@example.org> wrote:
>> Henrik wrote to me privately. I want to underline that this is
>> deprecated as a way of continuing Statalist threads for all the
>> reasons already explained in detail at
>> Indeed, here's another reason not mentioned there, but I trust clear
>> enough. A posting to a Statalist thread can often contain precisely
>> all that someone knows about a subject. It is not an offer to provide
>> further private support.
>> Now for a direct answer.
>> Sorry, but as I already indicated I don't do this kind of thing and I
>> have no idea whether your problem has already been solved, or even if
>> it is a good question.
>> Also, I doubt that the manuals anywhere discuss propensity score
>> matching. This is one of many areas which StataCorp is, I guess, very
>> happy to leave to users who have that specific interest and do work in
>> that area.
>> I hope you get better replies from people who know about this.
>> On Tue, Sep 27, 2011 at 3:47 PM, Henrik Wiig <email@example.com> wrote:
>>> Hi Nick!
>>> Thanks for your quick answer.
>>> Unfortunately, I do not know much about dose-response either, but talking to seasoned stata users and impact analysists, they have actually never separated the households by subcategories of the observations. It seems like the problem is finding the error term through bootstrapping, which has to be done for each district and then aggregated in one way or another afterward.
>>> And I have searched the stata manual for a long time without success. This surprices me too, as I would think that splitting into the dataset into homogenous cultures and then compare households, would be the obvious thing to do.
>>> In exactly what command do you find this problem solved?
>>> Fra: firstname.lastname@example.org [email@example.com] på vegne av Nick Cox [firstname.lastname@example.org]
>>> I don't do this kind of thing at all, but
>>> . findit propensity
>>> indicates what has been done, and there is a lot. These are the
>>> examples to follow if your problem has not already been tackled.
>>> On Tue, Sep 27, 2011 at 3:06 PM, Henrik Wiig <email@example.com> wrote:
>>>> My cross section dataset is collected from 10 district with different local cultures and hence omitted variables that both affects my treatment and dependent variable. I would hence like to run Propensity Score Matching (PSM), but then only compare treatment with control households within each district, not between districts.
>>>> Does a ready made command in Stata exist? If not, someone gave me a tip of programming the operation in an .ado file. Have anyone done this and could hence send me an example of how this is done?
* For searches and help try: