[Limdep Nlogit List] RE: RPL question on specifying random parameters

Iain Pardoe ipardoe at lcbmail.uoregon.edu
Thu Jan 18 04:32:07 EST 2007


I received some helpful replies from Kenneth Train, David Hensher, Bill
Greene, and Maksym Polyakov (thanks to you all).  Here is some code that
fits the model in the Brownstone & Train and McFadden & Train papers:

NLOGIT ; Lhs = ch
       ; Choices = A1,A2,B1,B2,C1,C2
       ; Rhs = price,rang1,acc1,spee1,poll,size1,big,lugg,cost1,stav,
 
suv,spo,wag,tru,van,ev,noev,comev,colev,cng,nocng,meth,colmet
       ; RPL
       ; Fcn = noev(N,*,0),nocng(N,*,0),size1(N),lugg(N)
       ; Halton
       ; Pts = 100 $

Best wishes, Iain

-----Original Message-----
From: Iain Pardoe 
Sent: Tuesday, January 16, 2007 10:06 AM
To: 'Limdep and Nlogit Mailing List'
Subject: RPL question on specifying random parameters

While working some RPL/ML examples from the literature, I'm having some
trouble with the alternative vehicle application in Brownstone & Train
(J. Econometrics, 1999) and McFadden & Train (J. App. Econometrics,
2000).  In particular, for reasons I don't quite understand, two of the
four random effects that are used are associated with dummy variables
that are the reverse of the dummy variables used in the "means" part of
the model.  To illustrate, here is code for a model similar to those in
the papers but with both dummy variables in question ("ev" and "cng")
entering both parts of the model in the same way:

read;file="C:\Documents and Settings\ipardoe\My
Documents\limdep\misc\mt.txt";
nobs=27924;nvar=26;names=price,range,acc,speed,poll,size,big,lugg,cost,s
tav,
suv,spo,wag,tru,van,ev,comev,colev,cng,meth,colmet,nous1,ch,noev,nocng,n
ous2$

create ; rang1 = range/100
       ; acc1 = acc/10
       ; spee1 = speed/100
       ; size1 = size/10
       ; cost1 = cost/10 $

NLOGIT ; Lhs = ch
       ; Choices = A1,A2,B1,B2,C1,C2 ?generic design
       ; Rhs = price,rang1,acc1,spee1,poll,size1,big,lugg,cost1,stav,
               suv,spo,wag,tru,van,ev,comev,colev,cng,meth,colmet
       ; RPL
       ; Fcn = ev(N),cng(N),size1(N),lugg(N)
       ; Halton
       ; Pts = 100 $

However, the model in the papers appears to use "noev" (1-ev) and
"nocng" (1-cng) as random parameters (but fixing their means at zero and
leaving "ev" and "cng" in the "means" part of the model):

NLOGIT ; Lhs = ch
       ; Choices = A1,A2,B1,B2,C1,C2
       ; Rhs = price,rang1,acc1,spee1,poll,size1,big,lugg,cost1,stav,
               suv,spo,wag,tru,van,ev,comev,colev,cng,meth,colmet
       ; RPL
       ; Fcn = noev(N),nocng(N),size1(N),lugg(N)
       ; Halton
       ; Pts = 10 $

Clearly this doesn't work (and gives an error message) because "noev"
and "nocng" are not Rhs variables.

I'd be grateful for any insights anyone can provide.  Am I
misunderstanding the model in the B&T and M&T papers?  If not, is there
a way to fit this model with nlogit?

Best wishes, Iain

Iain Pardoe <ipardoe at lcbmail.uoregon.edu>
Assistant Professor of Decision Sciences
Charles H. Lundquist College of Business
1208 University of Oregon
Eugene, OR 97403-1208, USA
Ph: 541-346-3250, Fax: 541-346-3341
http://lcb1.uoregon.edu/ipardoe



More information about the Limdep mailing list