Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

st: RE: Diagnostic results of ivreg2

From   "Schaffer, Mark E" <[email protected]>
To   <[email protected]>
Subject   st: RE: Diagnostic results of ivreg2
Date   Thu, 29 Mar 2012 00:13:00 +0100


Can you explain what you mean when you say "However the test of
endogeneity fails"?  Do you mean that you fail to reject the null that
the regressor is exogenous?

That just means that there's no significant difference between your IV
and OLS estimates of the coefficient.  A reason to be careful here is if
the IV estimate is not very precise, i.e., the SE is large, and you have
good prior reasons to think that the coefficient is endogenous.  If
that's the case, then you may well be better off with the imprecise but
consistent estimator (IV) than with the (apparently) precise but
inconsistent estimator.  On the other hand, you could probably also
construct an argument that goes the other way on MSE or similar grounds,
i.e., you're willing to accept some (asymptotic) bias in exchange for an
estimator with a smaller variance.


> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of 
> Abekah Nkrumah
> Sent: 28 March 2012 13:47
> To: [email protected]
> Subject: st: Diagnostic results of ivreg2
> Hi All,
> I am running an IV for a continuous dependent variable. The 
> endogenous variable is also continuous so am using the 
> use-written ivreg2 command and I correct for intra-cluster 
> correlation by using the cluster robust option. My 
> first-stage results suggest that the following diagnostic 
> tests are fine: (1) underidentification test via Kleibergen 
> Paap Wald statistsic  (2) overidentification test via the 
> Hansen J statistic (3) weak instrument test via the 
> Crag-Donald F-statistic and the Stock-Yogo relative bias 
> thresholds (4) The F statistic for the joint significance of 
> the instruments in the first stage is also more than 10 for 
> purposes of Staiger and Stoke's  rule (5) Weak instrument 
> robust test via the Anderson -Rubin wald test.
> However the test of endogeneity fails, suggesting that the 
> endogenous regressor is not endogenous. I am however 
> optimistic from a theoretical perceptive that the purported 
> endogenous regressor is indeed endogenous. I am not too sure 
> what may be happening? Could this be taken as symptoms of 
> weak or invalid instruments even though it passes all the 
> relevant  statistical test? I will appreciate your help
> Gordon
> *
> *   For searches and help try:
> *
> *
> *

Heriot-Watt University is the Sunday Times
Scottish University of the Year 2011-2012

We invite research leaders and ambitious early career
researchers to join us in leading and driving research
in key inter-disciplinary themes. Please see

for further information and how to apply.

Heriot-Watt University is a Scottish charity
registered under charity number SC000278.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index