[Limdep Nlogit List] two-step nested logit estimation help

Kandice L. Kleiber kleiberk at onid.orst.edu
Sat Apr 4 12:28:51 EST 2009

Hi ... Because I only have access to LimDep (without NLOGIT) for the time, I
am trying to do a two-step estimation of a nested logit.  However, I'm not
sure how to set up my code for LimDep (rather than NLOGIT).

My model looks like:
tree = land[corn(cc, sc, wc, oc), soy, wheat]  where cc, sc, wc, oc are 0-1
variables (where previous land came from) and corn, soy, and wheat are
proportion data.

I first try to estimate the corn(cc, sc, wc, oc) and save the inclusive
values.  I've tried just the basic multinomial logit model ... but then I'm
not sure how to get my inclusive values into the next estimation set ... I
can't find anything that really talks about setting this up in the limdep
First step:
lhs = prop;
choices = cc, sc, wc, oc;
rhs = one, pr41, pr42, pr43, ind1, ind2, ind3, ... ;
ivb = incl0$

Second step:
lhs = corn, soy, wheat;
choices = corn, soy, wheat;
rhs = one, age1, age2, ...  $
*How do I fit in the inclusive value??

The other option I have tried is setting the model up using the nested logit
choices = cc, sc, wc, oc;
ivb = incl0;
tree = land[corn(cc, sc, wc, oc), soy, wheat];
model: U(cc) = bcc + bpr41*pr41+bind1*ind1 ... /
           U(sc) = bsc +.../
           U(wc) = bwc +... /
          U(oc) =       bpr44*pr44+ ... /
        U(corn) = ac+ ... /
        U(soy) = as+ ... /
        U(wheat) =     aed3*ed3 + ....;
ivb = incl;

But this won't run and I'm not sure LimDep supports this.  Help?  And thank
Kandice Kleiber

More information about the Limdep mailing list