[Limdep Nlogit List] Contents of Limdep digest: Different Betas between Output window and B-Matrix, and Warning 344

[Richard Turner]

Greetings,

I've estimated a discrete choice model with 260 parameters and at the end
of the model the following warning is shown: Warning  344: LOGIT/Fixed
effects. Cannot compute P.(Sum(yi) near 0 or T).

What does this warning mean?

Also, according to page 398 of ACA 2nd ed., using the NLOGIT command, you
can estimate models up to 300 parameters. When I compared the parameters
greater than 150 from the output window and the B-matrix, they are
different. All of the parameters after the 150th from the B-matrix are very
close to 0, while the parameters after the 150th from the output window
seem to be believable values. Which output of the parameters can be
trusted? Just to note, the parameters up to and including 150 are the same
though.

Could I be using the NLOGT command incorrectly? Blow is a template of the
model syntax I am using:

Nlogit
;Lhs=Choice,CSET,ALTID
;Choices=A1,A2,A3,A4,A5,A6,A7
;Rhs=one,
?260 parameters
$

Do I perhaps need to explicitly create utility functions? Is the difference
in the parameters related to the 344 warning?

Thank you very much for any help.

Regards,

R