[Limdep Nlogit List] V/v: Nlogit without differing attributes

[Mausam Duggal]

Dear Klandice,

Just to add a couple of points to Chinh's comments:

I think the default for NLOGIT is normalization at RU1, so technically you can avoid adding that in the syntax.

Further, it is possible to also test a model that has an upper level (CS) constant, but it is not necessary as Chinh suggests.

Lastly, the p-value reported for the IV parameter in the NLOGIT output indicates whether it is significantly different than ZERO. Thus, even if you do get an IV parameter value that is lesser than 1.0, you would be better of testing whether it is significantly different than ONE.

Good Luck!

Mausam Duggal, MCIP, RPP

Senior Project Coordinator

MMM GROUP

100 Commerce Valley Drive West

Thornhill, ON Canada L3T 0A1

T: 905-882-4211, x6289

F: 905-882-7277

W: www.mmm.ca


-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au [mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of ho quoc Chinh
Sent: Wednesday, June 24, 2009 8:37 PM
To: Limdep and Nlogit Mailing List
Subject: [Limdep Nlogit List] V/v: Nlogit without differing attributes

Dear Kandice.

It does not matter if you just have explanatory variables which are similar across alternatives. The model normally works as long as your specification is suitable with nest logit theory, that is each nest  must be seen as a MNL model. Then you should set at least one of elemental alternatives (corn, soy, wheat, other) equal to zero; all of the remaining alternatives are specified as normal. In your problem, I think the following specification is reasonable:

Nlogit;
start=logit;
RU1;
model:
U(corn)=asc_c+...../
U(wheat)=asc_w+.../
U(other)=asc_o+....$

It's no need to specify utility function at higher level (branch level, that is U(cs)). The probability of choosing higher level is already concerned in the structure of Nest logit model. What you have to take into consideration is whether or not IV parameter of your tree is larger than 1.0, which is neccessary to satisfy random utility theory.

By the way, I think that you misunderstood the meaning of "normalization" in NL model. It means that we normalize the IV of lower level or upper level to 1.0 instead of setting one of the alternatives in each level to zero as you though.

Best regards,

Chinh.
Nagoya University, Department of Civil Engineering.
Transportation Planning and Engineering lab.

--- Thứ 5, 25/06/09, Kandice L. Kleiber <kleiberk at onid.orst.edu> đã viết:

Từ: Kandice L. Kleiber <kleiberk at onid.orst.edu>
Chủ đề: [Limdep Nlogit List] Nlogit without differing attributes
Đến: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Ngày: Thứ Năm, 25 tháng 6, 2009, 4:26

Hi -

I'm trying to set up a nested logit.  My data, however, does not  
contain any variables that differ across choices.  My tree = cs (corn,  
soy), wheat, other.  The NLogit manual provides examples where some  
variables differ across choices, however this set-up doesn't work when  
all of the variables have the same value across differing choices.

Is it reasonable to set up the model excluding the choices that I want  
to normalize on?  For example, if I set up my model following the  
NLogit manual, I would have:
Model:
U(corn)
U(soy)
U(cs)
U(wheat)
U(other)

Since my variables do not vary across choices, can I instead set up  
like:
Model:
U(corn)
U(cs)
U(wheat)

Thus normalizing on the lower level with soy, and on the upper level  
with other.

Any input is appreciated.  Thank you!
Kandice.

---
Kandice L. Kleiber
Graduate Teaching Assistant
Department of Economics
Oregon State University
e: kleiberk at onid.orst.edu
w: 541.737.6628

_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au



      &quot;Tốt hơn, thoáng gọn hơn, nhanh hơn -Trải nghiệm Yahoo! Mail mới hôm nay!
http://vn.mail.yahoo.com&quot;
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au