Testing for correlations in a random effects model, i,e. testing Corr (Xi uit)=0

[Mohamud Hussein]

Hi members,

I would like to test for correlations between a set of explanatory variable and idiosyncratic error component of a random effects model, i.e. ascertain whether corr(xi, uit)=0.

My command line is as following:

 xtreg price_ad i.b1.cool branv packz prodc prodt m retai, re i(SKUCode)

Can you please tell me how I can do this test as postestimation?

Thanks,
Mohamud

[Carlo Lazzaro]

Mohamud:
-help xtreg_postestimation- will take you to -xttest0-, that might be what you're looking for.

[Mohamud Hussein]

Thanks Carlo.

xttest0 returned aBreusch and Pagan Lagrangian multiplier test for random effects.The null hypothesis for this test is var(ui)=0. I got a highly significant results meaning, I think, random effects model is more appropriate than OLs in the case.

However, I am not though sure of how this results are related to my to my intitial quest on testing for corr (xi, ui)=0? Can you to please tell me how the two issues are linked?

Thanks,
Mohamud