Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: RE: Hausman Test Problems


From   John Antonakis <John.Antonakis@unil.ch>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: Hausman Test Problems
Date   Mon, 02 May 2011 12:34:11 +0200

Right; thanks for the clarification, Mark. Maybe I was not clear enough, but if one model is nested in the other, then one makes a restriction (that the L2 predictors do not correlate with the constant effect), which gives the overid interpretation. The reason why I talked about this is that I thought that Muhammad was thinking that he need to have instruments (where he said that he did not "pretest it for overidentification").

BTW, the paper I cited below is a real eye opener because it show just how many restrictions are in these models (for those who are interested):

Bollen, K. A.,&  Brand, J. E. (2010). A General Panel Model with Random and Fixed Effects A Structural Equations
Approach. Social Forces, 89(1), 1-34.

Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 02.05.2011 11:45, Schaffer, Mark E wrote:
John, Muhammad,

The test of fixed vs. random effects (also) has an overid test
interpretation.  The FE estimator uses the moment conditions
E(x_it*e_it)=0.  The RE estimator uses, in addition, the moment
conditions E(x_it*u_i)=0.  That's what makes it overidentified and an
overid test possible.

There is a short discussion and some references in the xtoverid help
file.

Cheers,
Mark

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of
John Antonakis
Sent: 02 May 2011 10:11
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: RE: Hausman Test Problems

You don't need to be overidentified to use xtoverid.  The
command in fact tests a constraint that is made, which nests
the random and fixed-effects models (i.e., the constraint
that is made to the random effects model is that level 2
regressors do not correate with uj).

To get a better handle on what types of constraints are made
in these types of models see:

Bollen, K. A.,&  Brand, J. E. (2010). A General Panel Model
with Random and Fixed Effects A Structural Equations
Approach. Social Forces, 89(1), 1-34.

HTH,
John.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 02.05.2011 10:56, Muhammad Anees wrote:
Thanks Eric!

It did worked for me. I actually run the regressions without
pretesting it for any overidentification. Can I still follow any
procedure selecting one of the FE and RE using over
identified panel
data regressions.

On 2 May 2011 12:44, DE SOUZA
Eric<eric.de_souza@coleurope.eu>   wrote:
The Hausman test for fixed vs  random is only valid under
a strict set of assumptions. These assumptions are clearly
not satisfied in your case .
Use -xtoverid-. Download it from ssc: -ssc install
xtoverid- and read the help file first.

Eric de Souza
College of Europe
Brugge (Bruges), Belgium
http://www.coleurope.eu


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of
Muhammad
Anees
Sent: 02 May 2011 06:12
To: statalist@hsphsun2.harvard.edu
Subject: st: Hausman Test Problems

Dear All!

I have run a panel data regression and selection of the
random effects or fixed effects using Hausman test. I do not
know what is the actual problem with my results. Please could
someone help. Why the result for my hausman command results
in warning message?
the complete results are below:


. xtreg priceclose eps bookvalue, fe

Fixed-effects (within) regression               Number of
obs      =       850
Group variable: id                              Number of
groups   =       170
R-sq:  within  = 0.1160                         Obs per
group: min =         5
between = 0.5266
avg =       5.0
overall = 0.4645
max =         5
F(2,678)           =     44.48
corr(u_i, Xb)  = 0.4836                         Prob>   F
         =    0.0000

priceclose       Coef.   Std. Err.      t    P>t     [95%
Conf. Interval]
eps    .7770481   .1966364     3.95   0.000     .3909585
  1.163138
bookvalue    .8653121   .1577343     5.49   0.000
.5556057    1.175018
_cons    1.001173   .1176642     8.51   0.000     .7701434
    1.232204
sigma_u   3.5662704
sigma_e   1.5953308
rho   .83325562   (fraction of variance due to u_i)

F test that all u_i=0:     F(169, 678) =    17.34
   Prob>   F = 0.0000
.
. estimates store fe

.
. xtreg priceclose eps bookvalue, re

Random-effects GLS regression                   Number of
obs      =       850
Group variable: id                              Number of
groups   =       170
R-sq:  within  = 0.1159                         Obs per
group: min =         5
between = 0.5186
avg =       5.0
overall = 0.4593
max =         5
Random effects u_i ~ Gaussian                   Wald
chi2(2)       =    297.79
corr(u_i, X)       = 0 (assumed)                Prob>
chi2        =    0.0000

priceclose       Coef.   Std. Err.      z    P>z     [95%
Conf. Interval]
eps    1.113035   .2084971     5.34   0.000     .7043883
  1.521682
bookvalue    1.394302   .1196459    11.65   0.000
1.159801    1.628804
_cons    .5629992   .2070207     2.72   0.007     .1572462
    .9687522
sigma_u   2.1242726
sigma_e   1.5953308
rho   .63938518   (fraction of variance due to u_i)


.
. estimates store re

.
. hausman fe re

---- Coefficients ----
(b)          (B)            (b-B)     sqrt(diag(V_b-V_B))
fe           re         Difference          S.E.

eps     .7770481     1.113035       -.3359869               .
bookvalue     .8653121     1.394302       -.5289903         .102786

b = consistent under Ho and Ha; obtained from xtreg B =
inconsistent
under Ha, efficient under Ho; obtained from xtreg

Test:  Ho:  difference in coefficients not systematic

chi2(2) = (b-B)'[(V_b-V_B)^(-1)](b-B)
=   -15.59    chi2<0 ==>   model fitted on these
data fails to meet the asymptotic
assumptions of the Hausman test;
see suest for a generalized test


--
Muhammad Anees
MSc in Economics
The University of Sheffield
United Kingdom
*
*   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/


*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index