hausman test correlated random effects
Moderators: EViews Gareth, EViews Moderator
hausman test correlated random effects
After estimating my model with random effects specification, I run the correlated random effects  hausman test, but results include the warning "crosssection test variance is invalid. Hausman statistic set to zero." Also, "robust standard errors may not be consistent with assumptions of Hausman test variance calculation." Have I done something wrong in the initial estimation or how do I interpret the test results in light of these warnings?

 EViews Developer
 Posts: 2600
 Joined: Wed Oct 15, 2008 9:17 am
Re: hausman test correlated random effects
The Hausman test statistic estimator for the variance of the coefficient difference is not guaranteed to be positive definite. In cases where one cannot compute the statistic, we set the value equal to zero and report the first warning message that you encountered.
The robust message is generated whenever you are estimating using robust standard errors. In this setting, you are saying that you believe that the standard model is not efficient, which is a requirement for the computation of the Hausman test variance.
The robust message is generated whenever you are estimating using robust standard errors. In this setting, you are saying that you believe that the standard model is not efficient, which is a requirement for the computation of the Hausman test variance.
Re: hausman test correlated random effects
Hallo glenn, i am a new user of eviews and i really need need your help to fix my problem on my final undergraduate thesis, so i have some similiar problem with hausman test like this, would you like to explain me with very simple explanation about the warning "crosssection test variance is invalid. Hausman statistic set to zero." and also "robust standard errors may not be consistent with assumptions of Hausman test variance calculation."
if the data I am working on has a big error or can still run?
I really need your confirmation about this please... Thank you
if the data I am working on has a big error or can still run?
I really need your confirmation about this please... Thank you
Last edited by April05 on Sun Feb 18, 2018 1:15 pm, edited 1 time in total.

 EViews Developer
 Posts: 2600
 Joined: Wed Oct 15, 2008 9:17 am
Re: hausman test correlated random effects
There's nothing wrong with the estimation, per se.
The zero variance statement means that the estimator of the random effects variance is zero. In this case, the RE estimates are the same as OLS.
The robust message is saying that you are estimating the coefficients using robust standard errors but that the test that you are performing is not robust. It's just an informative message.
The zero variance statement means that the estimator of the random effects variance is zero. In this case, the RE estimates are the same as OLS.
The robust message is saying that you are estimating the coefficients using robust standard errors but that the test that you are performing is not robust. It's just an informative message.
Re: hausman test correlated random effects
I really say thanks for your answer glen, your explanation help me to fix my problem...
And one more question about this glenn, if my hausman test condition was 1.00 and hausman statistic set to zero. Does it mean that i have to use Fixed Effect Model?
Thank you
And one more question about this glenn, if my hausman test condition was 1.00 and hausman statistic set to zero. Does it mean that i have to use Fixed Effect Model?
Thank you

 EViews Developer
 Posts: 2600
 Joined: Wed Oct 15, 2008 9:17 am
Re: hausman test correlated random effects
If your estimate of the residual variance is zero, it says that the RE model indicates you don't have individual effects. But that estimate may be problematic if the effects are correlated with regressors.
The fixed effects estimator is consistent even if there is correlation so it's a safer, though less efficient estimator. That's the basis of the Hausman test.
The fixed effects estimator is consistent even if there is correlation so it's a safer, though less efficient estimator. That's the basis of the Hausman test.
Return to “Econometric Discussions”
Who is online
Users browsing this forum: No registered users and 7 guests