[Limdep Nlogit List] Non-linear optimization in Limdep

[Angel Bujosa Bestard]

Thanks for all your comments, they have been very helpfull!!

I would like to make another question about the code send by Nanni Concu.
It looks like it is fitted for a model with three alternatives, doesn't it?
In that case, do you have the data in the usual limdep format, or dou you have
it arranged in another way? It looks to me that in each observation (row) you
have all the variables (attributes) related to all alternatives. I mean, are W1, W2
and W3 the values of the attribute WEED for three alternatives of the model?

Usually, I use my data in the traditional limdep format, one row by individual and
alternative. Then I only have one variable for each choice attribute, which takes
different values in each row. Do I need to change the arrengement of my data?

Again, many thanks!!! 

A. Bujosa



----- Original Message ----- 
To: <limdep at limdep.itls.usyd.edu.au <http://limdep.itls.usyd.edu.au/mailman/listinfo/limdep>>
Subject: [Limdep Nlogit List] Non-linear optimization in Limdep


 A.Bujosa,
I wrote this code lond ago, and don't remember all the details now.
ASC* are Alternative Specific constants; ASC_* are interactions with 
individual characteristics; WEED, FIRE, ACCS and COST are choice attributes.
I think the part that should interest you is the bit in the Fnc 
specification. As you can see is a simple Conditional Logit. Hope it helps.

NAMELIST;x=ASC1,A1AT,A1INC,AS1,AED11,AED12,AED13,AI1,W1,F1,N1,C1 $
NAMELIST;y=ASC2,A2AT,A2INC,AS2,AED21,AED22,AED23,AI2,W2,F2,N2,C2 $
NAMELIST;z=ASC3,A3AT,A3INC,AS3,AED31,AED32,AED33,AI3,W3,F3,N3,C3 $
MAXIMIZE
;labels=ASC,ASC_ATT, ASC_INC, ASC_SEX,A_Y12, A_CERT, 
A_UNI,A_INFO,WEED,FIRE,ACCS,COST
;Fcn=
bx=(ASC'x)|
by=(ASC'y)|
bz=(ASC'z)|
bset1=Exp(bx)|
bset2=Exp(by)|
bset3=Exp(bz)|
bset=bset1+bset2+bset3|
pr1=bset1/bset|
pr2=bset2/bset|
pr3=bset3/bset|
d1*Log(pr1)+d2*Log(pr2)+d3*Log(pr3)
;start=0,0,0,0,0,0,0,0,0,0,0,0 $
CALC; nodist=logl$

Nanni Concu
Univeristy of Queensland


----- Original Message ----- 
From: "Angel Bujosa Bestard" <angel.bujosa at uib.es <http://limdep.itls.usyd.edu.au/mailman/listinfo/limdep>>
To: <limdep at limdep.itls.usyd.edu.au <http://limdep.itls.usyd.edu.au/mailman/listinfo/limdep>>
Sent: Monday, August 20, 2007 6:45 PM
Subject: [Limdep Nlogit List] Non-linear optimization in Limdep


>/I am wondering if it is possible to use Limdep-Nlogit for estimating a
/>/ variation of a conditional logit model. That is, it is possible to use
/>/ the minimize/maximize commands to optimize the likelihood function of a
/>/ conditional logit model (or a variation of it)?
/>/
/>/ In such case, does anybody has some examples??
/>/
/>/ Thanks!
/>/
/>/
/>/ --
/>/ A.Bujosa
/>/ Balearic Islands University
/>/
/>/ _______________________________________________
/>/ Limdep site list
/>/ Limdep at limdep.itls.usyd.edu.au <http://limdep.itls.usyd.edu.au/mailman/listinfo/limdep>
/>/ http://limdep.itls.usyd.edu.au
/>/ /