Date:         Wed, 8 Feb 2012 09:25:51 -0600
Reply-To:     Robin R High <rhigh@UNMC.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Robin R High <rhigh@UNMC.EDU>
Subject:      Re: Fixed effect regression with clustered standard errors, help!
Comments: To: k7br@GMX.FR
In-Reply-To:  <CAF78DZ0UzmLBUzB-EVUXdXF7JQntKdg+Q2DgNcfYbdWatJCDdg@mail.gmail.com>
Content-Type: text/plain; charset="US-ASCII"

I don't have time to look at the details now, but as a suggestion perhaps
could check the latter sections of Chapter 2 in Paul Allison's "Fixed
Effects" SAS book regarding the centered score model with random effects,
perhaps running the resulting dataset with HPMIXED instead of MIXED.

Robin High
UNMC






From:
SUBSCRIBE SAS-L Anonymous <k7br@GMX.FR>
To:
SAS-L@LISTSERV.UGA.EDU
Date:
02/06/2012 08:04 AM
Subject:
Re: Fixed effect regression with clustered standard errors, help!
Sent by:
"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>



Dear SAS list,

Many thanks for your suggestions... However it seems to me that these
procedures you suggested (proc Mixed or proc Surveyreg) allow indeed to
cluster the standard errors by individuals but do not absorb the
individual
dummies during the computation... My problem is that my PC cannot handle
the creation of these dummies (too many) and I should use the de-mean
mathematical trick to remove them from the estimation of the fixed effects
(not random effect) linear model
Are you eventually aware of any procedure to do that ?
again, the Stata counterpart would be areg Y X, absorb(id) vce(cluster id)
OR  or xtreg Y X, fe vce(cluster (id))

Many, many thanks again

On 6 February 2012 00:14, student statistics
<student.statistics@gmail.com>wrote:

> Just use proc mixed....:
>
> Proc mixed data=abc;
>
> model Y=X;
>
> repeated subject=id;
>
> run;
> On Feb 5, 2012 5:16 PM, "SUBSCRIBE SAS-L Anonymous" <k7br@gmx.fr> wrote:
>
>> Dear SAS-list,
>>
>> I have a very simple question :
>>
>> How can I estimate a fixed effect regression with clustered standard
>> errors
>> by individuals ? For those who also use Stata I would like to perform
the
>> same calculation as xtreg Y X, fe vce(cluster id).
>>
>> I cannot use the PROC PANEL because individuals in my database could
have
>> multiple observations on the SAME date...
>>
>> I have seen that proc GLM has an ABSORB option, but I cannot use
clustered
>> standard errors I think...
>> Moreover the number of individuals is such that I CANNOT create a dummy
>> variable for each individual (out of memory) so I must stick to absorb
>> procedures...
>>
>> Any idea ?
>>
>> Many, many thanks
>> Francesco
>>
>