Danny,
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of
Danny Cohen-Zada
Sent: 27 February 2008 20:39
To: [email protected]
Subject: Re: st: Critical values for stock and yogo test when
the clusteroption is used
Thank you very very much
I have one last question.
Actually, i am estimating an ivprobit model (with cluster)
and with one endogenous regressor and one excluded
instrument. Since I know there is no test for weak instrument
for ivprobit i run ivreg2 only to check that my instrument
are not weak (I thought this is the best test that i can
perform). I estimate ivprobit with maximum likelihood and not
with the two stage option. Thus, the first stage is slightly
different. Actually, in this estimation the f-statistic on
the exluded instrument is higher than that obtained with the
two stage procedure of ivreg. Should i compare the
f-statistic on the excluded instrument (from the maximum
likelyhood ivprobit
model) to the critical value, or should i take the
f-statistic from the first stage of the ivreg procedure?
I'm not aware of any studies of weak identification for ivprobit
estimation. Unless you can find something, you're on your own.
That said, Leandro Magnusson (now at Tulane) has done some interesting
work on weak identification with limited dependent variable models:
http://www.econ.brown.edu/students/Leandro%5FMagnusson/Research.html
Cheers,
Mark
Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3296
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes
Thanks again
Danny
----- Original Message -----
From: "Austin Nichols" <[email protected]>
To: <[email protected]>
Sent: Wednesday, February 27, 2008 7:50 PM
Subject: Re: st: Critical values for stock and yogo test when
the clusteroption is used
> Danny--
> This is your second question from yesterday, to which I answered
> (yesterday) that you should compare the cluster-robust statistic
> (5.26) to the crit value and therefore conclude you have a problem.
> You will find discussion of constructing Anderson Rubin confidence
> regions in http://www.stata.com/meeting/5nasug/wiv.pdf (and
its refs).
>
> If you strongly believe your model is correct:
> ivreg2 y1 x1 x2 (y2=z1), ffirst cluster (x3) you can also
add excluded
> instruments like so:
> g x1z1=x1*z1
> g x2z1=x2*z1
> g z2=z1^2
> ivreg2 y1 x1 x2 (y2=z1 x1z1 x2z1 z2), ffirst cue cl(x3) and now you
> have overID tests available to you as well.
>
> On Wed, Feb 27, 2008 at 12:22 PM, Danny Cohen-Zada
<[email protected]>
> wrote:
>> Dear Professor Shaffer,
>>
>> I thought again and i think i understood that i should
compare 5.26
>> to
>> 16.38
>> and i should be worry about the strength of my instrument.
Am i right?
>>
>> Danny
>>
> *
> * 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/
>
*
* 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/
*
* 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/
