Carlos
Can you say a bit more about what you are doing? My memory is that IV
are not really handled in an exploratory fashion as you seem to be
contemplating. It is true in order to make the the 2SlS "work"
better, you need some instruments that predict your endogenous
variable(s) with some power. However, when the program runs the IV
portion of the model, it throws everything that is exogenous,
including excluded exogenous variables, into the IV equation,
regardless of whether they are correlated with the endogenous
variables or not.
A good source on this is Hayduk's book in my opinion. More
challenging are Ken Bollen's books.
---
Paul Gronke Ph: 503-517-7393
Professor Fax:
Reed College
3203 SE Woodstock Blvd.
Portland OR 97202
http://www.reed.edu/~gronkep
On Dec 30, 2008, at 9:27 PM, Carlos Rodriguez wrote:
> Dear all,
>
> In order for an IV to have power, is it enough for it have a
> statistically significant correlation with the endogenous regressor or
> is there also some magnitude threshold for this coefficient? Would an
> IV that has a statiscally significant correlation with the endogenous
> regressor, but it's coefficient is just say 0.50, a valid strong
> instrument? Or would it be too weak?
>
> I read this can be checked by the reduced form regression (ie,
> regressing the dv of interest on the exogenous controls and the
> proposed IV) and then comparing if the coefficient for the latter is
> similar to the one obtained in the iv-regression (2sls, for example).
> Could anyone please clarify what "similar" means here? Is it enough
> if the have the same sign but former is barely significant and the
> latter is not?
>
> Thanks so much. I learn a lot from all the exchanges on the list.
>
> happy new year
>
> **********************************************************
> Political Methodology E-Mail List
> Editors: Melanie Goodrich, <[log in to unmask]>
> Delia Bailey, <[log in to unmask]>
> **********************************************************
> Send messages to [log in to unmask]
> To join the list, cancel your subscription, or modify
> your subscription settings visit:
>
> http://polmeth.wustl.edu/polmeth.php
>
> **********************************************************
>
**********************************************************
Political Methodology E-Mail List
Editors: Melanie Goodrich, <[log in to unmask]>
Delia Bailey, <[log in to unmask]>
**********************************************************
Send messages to [log in to unmask]
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
|