RE vs FE: Hausman test returns negative chi2.

[Karim Syed]

Dear Carlo and all other statalisters

I am encountering the same problem as mentioned by Amira in #1 (https://www.statalist.org/forums/for...chibar2-1-0000). SO I chose OLS over RE. But, when I run the hausman test to select between RE and FE models, the results shows :

Test: Ho: difference in coefficients not systematic
chi2(5) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= -2.18 chi2<0 ==> model fitted on these
data fails to meet the asymptotic
assumptions of the Hausman test;
see suest for a generalized test.

How should I proceed?

Thanks.

Karim

PS. Technically, if the RE model is already rejected in xttest0, why should we still apply hausman to choose between RE and FE? Shouldn't we apply some test to choose between pooled OLS and FE?

[Carlo Lazzaro]

Karim:
-try re-running -hausman- with -sigmamore- option;
- the test you mention is available as an F-test, appearing as a footnote to -xtreg, fe- outcome table. If the the F-test fails to reach statistical significance, pooled OLS outperforms -xtreg, fe-.

[Jesse Wursten]

Additionally, consider using xtoverid instead, which allows for clusterings. The standard hausman test assumes idiosyncratic errors (if you would use cluster/robust in your regressions, you should also use xtoverid, ... instead of hausman).

[Karim Syed]

Thanks Carlos and Jesse.
Best regards
Karim

[pritika dua]

Hello everyone,
I am analyzing a sample of panel data. Because of the count dependent variable I use xtnbreg and xtpoisson.I want to apply a Hausman test to choose between fixed and random effects. I face this problem. "Chi2<0 ==> model fitted on these data fails to meet the asymptotic assumptions of the Hausman test; see suest for a generalized test" In earlier questions I found in the Statlist archive the options 'sigmamore' or 'sigmaless' serve to solve this problem but when i run it i get: "Estimators do not save e(sigma) or e(rmse), sigma option not allowed" Also when i try xtoverid,i get:
xtoverid works only after xtreg, xtivreg, xtivreg2 or xthtaylor
last estimates not found.

Kindly guide me further

[Carlo Lazzaro]

Pritika:
-hausman- works well when a handful of assumptions are satifsied; unfortunately, this is often not the case.
You can try -suest- (but it won't work either if you do not have a score function for your regression models).
Probably, you received a negative chi2 from -hausman- (without -sigmamore- option): as reported in Example #2, -hausman- entry, Stata .pdf manual, you can interpret that result as you cannot reject the null hypothesis; then, go -re-, remembering, however, that -fe- in panel count models is condtional -fe- (that is, a beast that differs from the one we are familiar with under -xtreg-).

[Mustafa Can Kaya]

Hello mr. Lazzaro, i am new to Stata, what should i understand from ''then, go -re-, remembering, however, that -fe- in panel count models is condtional -fe- (that is, a beast that differs from the one we are familiar with under -xtreg-).''. Should i go with fe or re? Thank you Have a nice day Highest regards

[Carlo Lazzaro]

Mustafa:
welcome to this forum.
Go -xtlogit,re-.

[Nitin Jain]

Pritika:
-hausman- works well when a handful of assumptions are satifsied; unfortunately, this is often not the case.
You can try -suest- (but it won't work either if you do not have a score function for your regression models).
Probably, you received a negative chi2 from -hausman- (without -sigmamore- option): as reported in Example #2, -hausman- entry, Stata .pdf manual, you can interpret that result as you cannot reject the null hypothesis; then, go -re-, remembering, however, that -fe- in panel count models is condtional -fe- (that is, a beast that differs from the one we are familiar with under -xtreg-).

Hi, Carlo, I am facing a similar issue with xtnbreg. It gives the following output. Can you please elaborate what is a score function in your above reply.

hausman fe re

b = Consistent under H0 and Ha; obtained from xtnbreg.
B = Inconsistent under Ha, efficient under H0; obtained from xtnbreg.

Test of H0: Difference in coefficients not systematic

chi2(19) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= -46.89

Warning: chi2 < 0 ==> model fitted on these data
fails to meet the asymptotic assumptions
of the Hausman test; see suest for a
generalized test.

[Carlo Lazzaro]

Nitin:
what if you add the -sigmamore- or -sigmaless- option?

[Nitin Jain]

Nitin:
what if you add the -sigmamore- or -sigmaless- option?

Thanks, Carlo. When I add either sigmamore or sigmaless, the following message is displayed. Is this because I am using xtnbreg?

hausman fe re , sigmaless
Estimators do not save e(sigma) or e(rmse),
sigma option not allowed

hausman fe re , sigmamore
Estimators do not save e(sigma) or e(rmse),
sigma option not allowed

[Carlo Lazzaro]

Nitin:
ye, I think so.
You may want to take a look at:
1) https://www.stata.com/statalist/arch...msg00398.html;
2) https://www.stata.com/statalist/arch.../msg00441.html

[Nitin Jain]

Nitin:
ye, I think so.
You may want to take a look at:
1) https://www.stata.com/statalist/arch...msg00398.html;
2) https://www.stata.com/statalist/arch.../msg00441.html

Thanks, Carlo!