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 at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: missing std errors---xtivreg2


From   Abhimanyu Arora <abhimanyu.arora1987@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: missing std errors---xtivreg2
Date   Fri, 8 Jun 2012 15:21:09 +0200

Hello again
Just to follow up,
I discovered a few things that might facilitate a healthy discussion.
Vince Wiggins has got a post here
http://www.stata.com/statalist/archive/2003-06/msg00646.html, that
might be somewhat related.
But as I infer from his answer to Mark, there's to be 1 one and rest
zeros. But when I summarize my >2500 observations, I have 40% ones.
Also I do get standard error for this variable if I cluster it 1-way
(or I guess level would be more econometrically correct), instead of
2-way.
But I hope to understand this puzzling issue thanks to your cooperation
Cheers
Abhimanyu





Then I see that one-way clustering does


On Fri, Jun 8, 2012 at 2:35 PM, Abhimanyu Arora
<abhimanyu.arora1987@gmail.com> wrote:
> Dear statalist
> I am estimating a panel-fixed effects model with two-way clustering
> with one endogenous variable making use of Schaffer et al's -xtivreg2-
>
> . which xtivreg2
> c:\ado\plus\x\xtivreg2.ado
> *! xtivreg2 1.0.13 28Aug2011
> *! author mes
>
> May I request you to shed some light on a couple of issues.
>
> I am getting missing standard error for one included instrument. It is
> a binary variable.
> I get the following warning in my output indicating that this variable
> is problematic.
>
> Warning: estimated covariance matrix of moment conditions not of full rank.
>         model tests should be interpreted with caution.
> Possible causes:
>         number of clusters insufficient to calculate robust covariance matrix
>         singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
> partial option may address problem.
>
> However, since I would have liked to see the significance of the
> coefficient of this variable, is there a way to address this problem
> (inability to infer) without -partial-? I guess one might have to drop
> observations...but is it so?
>
>
> Another thing I need to clarify is that in this case while for weak
> identification test using the KP rk Wald F statistic (=AP F) rejects
> the null, I am not sure about the underidentification test---the first
> stage reports Kleibergen-Paap rk LM statistic as well as AP chi-sq and
> these two are meaning different conclusions for the null. For the
> cluster-robust case which among the two is preferred, perhaps I might
> have read too fast in the help file so as to miss it, so seek your
> thoughts/references on this.
>
> I could share the results if you ask.
>
> With best wishes
> Abhimanyu
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/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/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


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