Statalist


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

st: RE: _rivtest_ for waek intruments


From   "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: _rivtest_ for waek intruments
Date   Fri, 30 Oct 2009 16:06:33 -0000

Filipa,

To answer your question - no, that's not how the methods implemented by -rivtest- work.  This approach eschews point estimation entirely in favour of reporting confidence intervals.  You'll see that in your example below, you have 95% confidence intervals for the coefficient on your variable of interest (hasmigr_US).

--Mark

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Filipa de Castro
> Sent: Friday, October 30, 2009 1:59 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: _rivtest_ for waek intruments
> 
> Dear all,
> 
> Following Mark Schaeffer´s suggestion I am estimating an instrument
> regression with weak instruments using
> ivreg2, and using the post-estimation programme (rivtest). My output
> after running rivtest is like this:
> 
> ************
> xi: ivreg2 goschool edad indchild oldest child45 child6 lowsixfam
> edumam1 edumam2 edumam35 edumam68 edumam912 mommarried  madrejefefam
> ownhome  survfam  madremigfam smalltown divorcesep agricolaheadspouse
> ctapropheadspouse oladummy i.iden (hasmigr_US= instrpi* instrac2*) if
> sex==1   & edad_10_13==1
> 
> rivtest, ci
> 
> Weak instrument robust tests and confidence sets for    linear IV
> H0: beta[goschool:hasmigr_US] = 0
> 
> 
>          Test       Statistic                        p-value  
>          95%
> Confidence Set
> 
> CLR  stat(.)  =    64.31   Prob > stat =   0.0000
> [-1.75073,-.539857]
> AR  chi2(32) =   139.39   Prob > chi2 =   0.0000
> [-2.548374,-.2742049]
> LM  chi2(1)  =    39.41   Prob > chi2 =0.0000
> [-1.761058,-.5351977] U [ 1.438614, 2.739162]
> J  chi2(31) =      99.98   Prob > chi2 =   0.0000
> 
> LM-J              H0 rejected at 5% level
> 
> Wald  chi2(1)  =    32.29   Prob > chi2 =   0.0000
> [-.863005,-.420344]
> 
> Note: Wald test not robust to weak instruments. LM-J    
> confidence      set
> not available with closed-form estimation (use  usegrid option).
> ************
> 
> I wonder if you could help with the following question:
> I see that _rivtest_ does an hypothesis test in which Ho means that
> the endogenous variable coefficient is zero. In this particular case
> the coefficient of the endogenous variable is statistically different
> from zero. Considering that estimation with weak instruments could
> lead to biased estimates, is there a way to find the real level of the
> coefficient of the endogenous variable?
> 
> Many thanks
> Filipa de Castro
> 
> *
> *   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/
> 


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


*
*   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   |   What's new   |   Site index