Bookmark and Share

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: fixed effects quantile regression


From   Lisa Marie Yarnell <lisayarnell@yahoo.com>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: fixed effects quantile regression
Date   Wed, 12 Dec 2012 06:50:00 -0800 (PST)

Hi Maarten,

OK, maybe I should be trying the fixed effects quantile model using "qreg" instead of "sqreg"?  (I have seen this on various websites, so others have done it.)

If I want estimates for the 50th and 80th quantiles, can I run this using two separate xi: qreg models?  I am guessing that with enough reps, and perhaps repeated runs with different seeds, that reliable estimates for the 50th and 80th quantiles could be determined using two separate runs of the model instead of the simultaneous?

Thank you, 
Lisa

----- Original Message -----
One problem is that -sqreg- uses the bootstrap to calculate standard
errors, and in this case you'll probably want to draw individuals
rather than observations. The dropped variables are probably the
result of the bootstrap drawing "by accident" no observations for
those individuals. In terms of the Stata command -bootstrap- this
would mean that you'd probably have to specify the -cluster- and
-idcluster()- options and change your estimation command accordingly.
However, -sqreg- does not allow these options. So unfortunately the
answer is that this model is not implemented in Stata unless you
program it yourself.

Hope this helps,
Maarten




From: Maarten Buis <maartenlbuis@gmail.com>
To: statalist@hsphsun2.harvard.edu
Cc: 
Sent: Wednesday, December 12, 2012 12:40 AM
Subject: Re: st: fixed effects quantile regression

On Wed, Dec 12, 2012 at 2:34 AM, Lisa Marie Yarnell wrote:
> I am running the following quantile regression model, which attempts to model the id fixed effects by incuding the term i.id.
>
> Can anyone give me a clue about why the dummies represented by the i.id were dropped for reasons of multicollinearity?  Does it seem that I did not arrange my data properly, or did I specify the model incorrectly?  Are there any other ideas about what's going wrong here?
>
> The output was very long, showing all of the dropped predictors, so I shortened it below by writing "(cont'd)".
>
> Thanks in advance for the help,
> Lisa
>
> . xi: sqreg viol span otherlang viobeh_p20_b viobeh_p50_b viobeh_p80_b viobeh_p20_g viobeh_p50_g viobeh_p80_g seven eight i.id if gender==0, quantile (.5, .8) reps (200)
>
> i.id              _Iidc100001-50000055(naturally coded; _Iidc100001 omitted)
> note: _Iidc100004 dropped because of collinearity
> note: _Iidc100006 dropped because of collinearity
> note: _Iidc100007 dropped because of collinearity

*
*   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/


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