LR TEST in FINE and GREY model

[Mich Prov]

Dear all,
Having this syntax:

stset fu_esrd_2, fail(r_death==2) id(id_progressivo)
stcrreg eta2 sesso diabete bmi2 p2 hb2 i.stadi2 prot2 i.risk_cat pas2 delta_k, compete(r_death==1)
est store a
stcrreg i.k5_vis12 eta2 sesso diabete bmi2 p2 hb2 i.stadi2 prot2 i.ras1_2 pas2,compete(r_death==1)
est store b
lrtest a b

I saw that lrtest does not work because we obtained a log pseudolikelihood rather than a loglikelihood... so how can we do a model comparison (similar to lrtest from logistic regression)?

[Clyde Schechter]

Um, neither of these models is nested in the other. So I don't think you'd be able to do a likelihood-ratio test for this even if the models were actually estimated by real maximum likelihood.