Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: instrumental variable for quantile regression


From   "Austin Nichols" <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: instrumental variable for quantile regression
Date   Wed, 28 May 2008 09:39:36 -0400

alessia matano <alexis.rtd@gmail.com>:
I can't see where you got a repeated time values error.  As I said before:
"show an example using a publicly available
dataset, e.g. start your example code with
  webuse psidextract, clear
  tsset
Probably you need at least the idcluster(newvar) option on
-bootstrap-. "

This estimation procedure you are using is wandering further and
further afield; do you have any confidence it will generate consistent
estimates of the parameter you are interested in?  As I understand it,
you want to know the effect of some X on the conditional median of Y.
If I were you, I would see how your cluster-bootstrapped time-demeaned
qreg estimate does at recovering this parameter in some simulated
data.

On Wed, May 28, 2008 at 8:22 AM, alessia matano <alexis.rtd@gmail.com> wrote:
> Dear Austin.
>
> You are right I tried to demean the data myself, since when I used
> xtdata....., fe i(ind) it didn't create deviations from time average
> for each individual (ind). But, then I asked myself if it does not in
> fact do that, but just prepare the data for such a transoformation.
> Any help regarding this is welcome.
>
> Secondly i applied a bootstrap procedure like the following (after
> time demeaning the data)
>
> prog me
>  qreg dmwage dmindchar dmspec  dmd1995-dmd2004 dmsectors dmarea, q(0.1)
>  end
> bs, cluster(ind): me
>
> I before tsset the data
> tsset ind year
> and it recognize that it is a panel. Then when I apply the program he
> wrote repeated time values within panel. hence, my question about
> clustering.
> Moreover generally, in fixed effects estimates I also clustered the
> observations for area (to deal with heteroskedasticity questions), and
> I do not know if I can do it now.
> Last, I tried the same with clustering() (see below) and it stopped at
> the third bootstrap writing something like:
> ..x
>
> prog me
>  qreg dmwage dmindchar dmspec  dmd1995-dmd2004 dmsectors dmarea, q(0.1)
>  end
> bs, cluster(): me
>
> Thanks again for any help
> alessia
>
> 2008/5/27 Austin Nichols <austinnichols@gmail.com>:
>> alessia matano <alexis.rtd@gmail.com>:
>> Perhaps if you show what you typed, folks on the list can offer
>> advice.  Even better, show an example using a publicly available
>> dataset, e.g. start your example code with
>>
>> webuse psidextract, clear
>> tsset
>>
>> Probably you need at least the idcluster(newvar) option on
>> -bootstrap-.  How are you demeaning yourself?  Better to use an -xt-
>> commmand.
>>
>> On Tue, May 27, 2008 at 6:13 AM, alessia matano <alexis.rtd@gmail.com> wrote:
>>> Dear Austin,
>>>
>>> I attempt to do the estimation with bootstrap....after time-demeaning
>>> my observation. I have just a doubt. It is true that bootstrap allows
>>> the cluster option, what i do not understand well is for what should i
>>> clustered my observations to allow correct standard errors after
>>> time-demenaing myself the observations. generally i clustered them not
>>> for individual, but for provinces where they live (cause of
>>> heteroskedasticity). However bootstrap either using the individual
>>> cluster or using the province ones does not work, saying that there
>>> are time repeated values within the panel, while when I tsset them it
>>> understands the panel structure. May you help me with this also? It
>>> works only when I cluster() with no variables inside.
>>> May you suggest me something? Also to read about?
>>> thank you very much
>>> alessia
>>>
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



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