[Limdep Nlogit List] V/v: nested logit inclusive value restrictions

William Greene wgreene at stern.nyu.edu
Fri Jul 3 19:50:37 EST 2009

Dear Chinh and Kandice. Chinh, thanks for this excellent tutorial
on RU1 and Ru2. About the data type, NLOGIT inspects the dependent
variable before doing any estimation. It will automatically adjust
if it finds that the LHS variable is proportions. The difficult
case is when you have both proportions and ones and zeros in the
data - which you should avoid.  For the data set in which the
LHS variable is always a set of proportions, you do not need to
modify the command. The program will sort this out on its own and
adjust how it computes the log likelihood. You will see a statement
of this in the list of ancillary results that appears above the 
table of coefficient estimates.
/Bill Greene

----- Original Message -----
From: "ho quoc Chinh" <hoqchinh at yahoo.com.vn>
To: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Sent: Thursday, July 2, 2009 11:32:54 PM GMT -05:00 US/Canada Eastern
Subject: [Limdep Nlogit List] V/v: nested logit inclusive value restrictions

Dear Kandice Kleiber,
In principle, the specification of  "tree= cs(corn, soy), wh(wheat), ot(other)" and that of "tree = cs(corn, soy), wheat, other" are almost the same. Nevertheless, the former is often used in NLOGIT reference guide.
NLOGIT will automatically recognize that your tree structure is partial degeneracy if there exist at least one nest with only one elemental alternative. For clearly understanding the differences between RU1 and RU2, you should refer to the following articles:
1.Hensher, D. A. & Greene, W. H. (2002) Specification and Estimation of the Nested Logit Model: Alternative Normalisations. Transportation Research Part B: Methodological, Vol. 36, 1-17.
2.Bhat, C. R., Sen, S. & Eluru, N. (2009) The Impact of Demographics, Built Environment Attributes, Vehicle Characteristics, and Gasoline Prices on Household Vehicle Holdings and Use. Transportation Research Part B: Methodological, Vol. 43, 1-18.
If you use RU1 as a chosen nomalization, then you should set the IV values of degenerate nests to 1.0 by using the command: IVset:(wh,ot)=1.0;
On the other hand, if RU2 is used, it's no need to set the IV values to 1.0 since it's done automatically by Nlogit.
Though RU2 is prefered to be the 'standard' normalization, it's sometimes hard to converge, compared to RU1 specification.

RU3 is the same as RU2 as pointed out in the Hensher's and Greene's paper (the above article) as well as in NLOGIT reference guide (Chapter 8).
Finally, since your dependent variable is proportion, there should be some command specification to tell NLOGIT the model type (Discrete choice or Proportion or .....Frequency or Rank?) I did not find any direct command for this specification in the guide reference though there are some pictures in chapter 3 (N3-8) and chapter 5 (N5-2) show this option.
If you are still confuse after finish reading these above articles and my suggestions, contact me for more information. I will send you the Nlogit command for model used in Hunt's (2000) paper and Hensher's and Greene's (2002) paper. The specific example in these papers will help you to understand more about the way we specify Nested Logit Model.
Sincerely yours,
Nagoya Uninversity, Graduate School of Engineering
Transportation Engineering and Planning lab, Department of Civil Engineering.

--- Thứ 6, 3/07/09, Kandice L. Kleiber <kleiberk at onid.orst.edu> đã viết:

Từ: Kandice L. Kleiber <kleiberk at onid.orst.edu>
Chủ đề: [Limdep Nlogit List] nested logit inclusive value restrictions
Đến: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Ngày: Thứ Sáu, 3 tháng 7, 2009, 8:36

Hi -

I'm struggling with setting up my nested logit model.

Currently my tree = cs(corn, soy), wheat, other.
My dependent variable is a proportion, and my independent variables  
are all observation specific (do not vary across choices).

However, after doing some reading (Hunt, 2000), I think that I have  
not been using the restrictions that might need to be set in order to  
estimate the model correctly.

I'm wondering (and I'm not sure if LimDep reads this the same way) if  
I should change my tree structure to
tree= cs(corn, soy), wh(wheat), ot(other)
to indicate that my model is a partially degenerate model.  After  
reading Hunt (2000) (deals particularly with partial degeneracy) I  
think I need to know more about the normalization that is going on in  
the NLOGIT program.

I've come across the RU1, RU2, and RU3 commands.  Is one of the these  
the default for the NLogit program?  I think that I would want to use  
RU1 - what is an example of where someone would choose to use RU2 and  

Finally, I think I would need to set the restriction on the inclusive  
value parameters such that IV(wh)=IV(ot)=1 and then let LimDep  
estimate IV(cs).  LimDep won't restrict those parameters on it's  
own ... right?

Sorry for the confusion, but thank you for your help!

Kandice Kleiber

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

      Làm thế nào để luôn cảm thấy hạnh phúc trong cuộc sống? 
Xem những ý kiến thú vị nhất tại Yahoo! Hỏi & Đáp.&quot; http://vn.answers.yahoo.com
Limdep site list
Limdep at limdep.itls.usyd.edu.au

More information about the Limdep mailing list