[Salem Abdussalam]

[QUO=Carlo Lazzaro;n1360824]Salem:
yes, in that instance the -fe- specification outperforms both -re- specification and (pooled) OLS.[/QUOTE]

Carlo, thank you so much for your reply. It is a belated thanks I know. But thank you very much indeed.

[Salem Abdussalam]

#4
QUO=Carlo Lazzaro;n1360824]Salem:
yes, in that instance the -fe- specification outperforms both -re- specification and (pooled) OLS.[/QUOTE]

Carlo, thank you so much for your reply. It is a belated thanks I know. But thank you very much indeed.

[kjiyengar]

I am a novice, so kindly bear with me.
I am getting the following results for Hausman test of FE v/s RE. Also, my Breusch Pagan Lagrangian Multiplier (LM) Test is reported below. I understand from Hausman Test that FE is better than RE and from LM test that RE is better than pooled OLS. Is my understanding correct and if so, can I just report FE results? Many thanks.

__________________________________________________ _____________
Test: Ho: difference in coefficients not systematic

chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 117.43
Prob>chi2 = 0.0000

__________________________________________________ ________________

xttest0

Breusch and Pagan Lagrangian multiplier test for random effects

lncash[tic1,t] = Xb + u[tic1] + e[tic1,t]

Estimated results:
| Var sd = sqrt(Var)
---------+-----------------------------
lncash | 1.522601 1.233937
e | .3265444 .5714407
u | .9121868 .9550847

Test: Var(u) = 0
chibar2(01) = 16582.01
Prob > chibar2 = 0.0000

[Carlo Lazzaro]

Yes, you're correct: -fe- specification seems the way to go.

[kjiyengar]

Thanks, Carlo.

[Pandelis Andreou]

I am so grateful for this discussion, cause I have the same situation:

I have 5 different panel data models and in all 5 the Breusch-Pagan tests result in:


Test: Var(u) = 0
chibar2(01) = 0.00
Prob > chibar2 = 1.0000


I know this is strong evidence that there are not random effects, but still the absolute numbers (1 and 0) look suspicious to me.

So, Jeff Wooldridge, you are suggesting that in any way RE does not fit I should go with pooled OLS anyway?

So, Dr. @Jeff Wooldridge,