[Limdep Nlogit List] Limdep Digest, Vol 190, Issue 1

[Richard Yao]

Dear Professor Greene:

I employed NLOGIT 6.0 to compute a latent class logit model with two
classes, as outlined below:

? LCL model - 2 latent classes
NLOGIT
    ;lhs = chosen,cset,altij
    ;choices=sq,a,b
    ;maxit=200 ? panel's length & Max. num. of iterations
    ;pds = times
    ;lcm ;pts = 2 ? specifies it is a LCM and the number of classes
    ;pds=times ? specifies the log-likelihood with the product
operator,taking the product of the probabilities from 6 choice sets
    ;parameters ? estimate individual specific parameters
    ;model:
     U(sq)=               beh1*e_h1 + beh2*e_h2 + bmn1*mn1 + bmn2*mn2 +
bpol1*pol1 + bpol2*pol2 + bcost*cost /
     U(a) = baltd*alt_d + beh1*e_h1 + beh2*e_h2 + bmn1*mn1 + bmn2*mn2 +
bpol1*pol1 + bpol2*pol2 + bcost*cost /
     U(b) = baltd*alt_d + beh1*e_h1 + beh2*e_h2 + bmn1*mn1 + bmn2*mn2 +
bpol1*pol1 + bpol2*pol2 + bcost*cost $

The execution was successful, yielding useful results. However, I observed
a comment in the MNL model estimates table:

"Warning: The model lacks a complete set of ASCs. R-squared is problematic.
Consider using the model setup with ;RHS=one to obtain LogL0."

I would appreciate your guidance on resolving this matter.

Thank you in advance for your response.

With best wishes,

Richard