Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: unbalanced panel+ SURE+IVREG2+Test Hausman


From   "Olivier Lohest" <ol@plan.be>
To   statalist@hsphsun2.harvard.edu
Subject   st: unbalanced panel+ SURE+IVREG2+Test Hausman
Date   Fri, 06 Oct 2006 11:38:31 +0100

I have an unbalanced panel with two period and I have to estimate two different equations with a lot of invariant time variable in my two equations

In a first step, i have estimated my equation by sample pooling (OLS) But I also would like to estimate by panel estimation

To address this dimension, two different regressions have been considered (i) fixed effect and (ii) random effect.

Several tests such as the Breusch-Pagan Lagrange multiplier (LM) test, Hausman specification test and the F-test were employed to choose the most appropriated model. For all equations, these tests reject the random effect model. (with the comman xtreg xttest0 and hausman test)

I know that the Breusch pagan developped in the procedure xttest0 (in stata 8.2) is robust if we have an unbalanced panel. This test reject the null hypothesis of the absence of an unoberserved effect

In a second step, I use the Hausman test in order to compare between the RE and the FE estimators, the test reject the null hypothesis => the fixed effect is more approapriate.

IS THIS TEST ROBUST IN THE PRESENCE OF AN UNBALANCED PANEL WITH SHORT TIME PERIOD DO YOU HAVE SOME SUGGESTIONS


iF this test is valid and if use the fixed effect model (within),  I will lost a lot of variable due to the presence of a loft of time invariant variables 
=> I prefer to estimate my model by pooling IS IT CORRECT DO YOU HAVE SOME SUGGESTION ?

In a third step, I have some problems of  endogeneity.  I assume that one of my variable (present in the two equations) may be  may be correlated with the error term and thus produce a biased coefficient.  

Considering the fact that it is not usefull to use a fixed effect model due to time invariant variables, I have estimated my two equation with the two least square procedure method on each pooled sample (unbalanced)   (procÚdure ivreg2)?

MAY I USE THE IVREG2 PROCEDURE TO DO THAT ?

On my pooled sample (which is unbalanced) may I test the heteroskedasticty and the endogoneity test proposed by the Ivreg2 procedure ? IS IT CORRECT on pooled sample and if have heteroskedastiicity ?

Considering this approach, the ivreg2 procedure with orthog reject the endogeneity of my potential endogenity variable on my pool sample

but if test the two equations with the ivreg2 command for each time periode, in one of my period I can reject th null hypothesis of endogeneity and in the second period (if have only two time period) I can not reject the null hypothesis => I have different results in terms of endogeneity on my pool sample ?

COULD YOU EXPLAIN THIS DIFFERENCE DO YOU HAVE SOME SUGGESTIONS

THANK YOUR VERY VERY MUCH FOR YOUR HELP

OLIVIER

i DONT FIND ANY RESPONSE IN WOOLDRIDGE OR BALTAGI 
 

**********************************************************************
Disclaimer: This e-mail may contain confidential information
which is intended only for the use of the recipient(s) named above.
If you have received this communication in error, please
notify the sender immediately and delete this e-mail from
your system.
Please note that e-mail messages cannot be considered as official
information from the Federal Planning Bureau.
**********************************************************************

 

**********************************************************************
Disclaimer: This e-mail may contain confidential information
which is intended only for the use of the recipient(s) named above.
If you have received this communication in error, please
notify the sender immediately and delete this e-mail from
your system.
Please note that e-mail messages cannot be considered as official
information from the Federal Planning Bureau.
**********************************************************************
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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