Re: st: Re: question on XTOVERID

[Davillas Apostolos <davillas@upatras.gr>]

<>

I am teribly sorry, I did not know that. 
Thanks for your help. 


On Mon, 22 Mar 2010 14:17:23 -0400, Christopher Baum <baum@bc.edu> wrote:
> <>
> Statalist protocol states that you should reply to the list, and not to
> the poster who responded to your question.
> 
> FEs allow X to be corr with unit-specific error, but not idiosyncratic
> error. If there is corr (X,epsilon) then FE is inconsistent (as it is
OLS
> wi/dummies) and you need xtivreg to get consistent ests. Thus treating
> xtreg,FE as consistent is an assumption which should be tested before
> treating it as the always-consistent alternative to xthtaylor.
> 
> From Stata manual, -xthtaylor- -- see last line
> 
> Description
> xthtaylor fits panel-data random-effects models in which some of the
> covariates are correlated
> with the unobserved individual-level random effect. The estimators,
> originally proposed by Hausman
> and Taylor (1981) and by Amemiya and MaCurdy (1986), are based on
> instrumental variables. By
> default, xthtaylor uses the Hausman–Taylor estimator. When the amacurdy
> option is specified,
> xthtaylor uses the Amemiya–MaCurdy estimator.
> Although the estimators implemented in xthtaylor and xtivreg (see [XT]
> xtivreg) use the
> method of instrumental variables, each command is designed for different
> problems. The estimators
> implemented in xtivreg assume that a subset of the explanatory variables
> in the model are correlated
> with the idiosyncratic error {epsilon}it. In contrast, the
Hausman–Taylor
> and Amemiya–MaCurdy estimators
> that are implemented in xthtaylor assume that some of the explanatory
> variables are correlated
> with the individual-level random effects, {u}i, but that none of the
> explanatory variables are correlated
> with the idiosyncratic error, {epslion}it.
> 
> Kit Baum   |   Boston College Economics and DIW Berlin   |  
> http://ideas.repec.org/e/pba1.html
> An Introduction to Stata Programming   |  
> http://www.stata-press.com/books/isp.html
> An Introduction to Modern Econometrics Using Stata   |  
> http://www.stata-press.com/books/imeus.html
> 
> On Mar 22, 2010, at 1:20 PM, Davillas Apostolos wrote:
> 
>> Dear Christopher F Baum,
>> 
>> Thanks for you reply in my question (statalist). 
>> Lets have:
>> 
>> y=a+zi*g+ xit*b+mi+eit
>> 
>> where zi, mi(error) time invariant 
>> and xit, eit(error) time variant. 
>> 
>> F.E. estimators require E(eit|zi, xit.....mi)=0  (I) and 
>> the Hausman-Taylor estimation require (I) and the exogenous variables
>> (say
>> x1it and z1i)to be uncorrelated with mi. 
>> 
>> Since, FE provide consistent estimators under weaker assumptions we can
>> apply as simple hausman test to compare between fe and H-T. If we do
not
>> reject H0, no significant difference between these coefficients, means
>> that
>> H-T provide consistent estimates. 
>> 
>> The -htoverid option tets the assumption (I), i.e. that all the
variables
>> are independent of the eit.       
>> However, I am not very sure that this is the case in Baltani's
text-book
>> (page 127, equation 7.46).
>> 
>> 
>> 
>> Apostolos Davillas
>> 
>> PhD Candidate
>> University of Patras
>> Department of Economics
>> University Campus, Rio
>> Tel.:(+30)6974427234
>> E-mail: davillas@upatras.gr
>> 
>> 
>> 
>> 
>> 
> 
> 
> *
> *   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/