Dear Statalist,

I am trying to figure out if it is necessary to include a random intercept in a three level model (time nested in firms nested in sectors). First, I estimate a model without random intercepts and compare it with a model with random intercept (and random coefficient of a time-varying firm level variable) for firms. Thereafter, I compare the latter model with another that also include a random intercept for sectors. The problem is that when trying to do the LR test for the two last models, the test assume that sector-level model is nested in firm-level model (which cannot be), and thus, do not give the test (see below).
Is there anything wrong in my approach?
I think the problem has to do with the vce(robust) error. Can I perform the same but without robust errors (even though the reported results will be for the model with robust error but using the LR test for the no robust model to check for the inclusion of random intercept)?

Thanks a lot in advance.

Code:
.
. lrtest logit logit_f, force

Likelihood-ratio test                                 LR chi2(3)  =   7938.01
(Assumption: logit nested in logit_f)                 Prob > chi2 =    0.0000

. local Likelihood_ratio_test1 = r(chi2)

. local pvalue_Likelihood_ratio_test1 = r(p)

.
. lrtest logit_f logit_s, force

Likelihood-ratio test                                 LR chi2(0)  =   -186.77
(Assumption: logit_s nested in logit_f)               Prob > chi2 =         .
Look when I do not use vce(robust) there is not problem with the tests
Code:
 lrtest logit logit_f, force

Likelihood-ratio test                                 LR chi2(3)  =   7938.01
(Assumption: logit nested in logit_f)                 Prob > chi2 =    0.0000

. local Likelihood_ratio_test1 = r(chi2)

. local pvalue_Likelihood_ratio_test1 = r(p)

.
. lrtest logit_f logit_s, force

Likelihood-ratio test                                 LR chi2(1)  =    186.77
(Assumption: logit_f nested in logit_s)               Prob > chi2 =    0.0000