[Limdep Nlogit List] Choice Model Design Software

[Nadhem Mtimet]

Dear John and Ron,

Sorry to respond you so late. John, I completely agree with you about the
fact that orthogonality is not a very important in constructing discrete
choice experiment (DCE) since that, as you said, orthogonality is desirable
in linear models and academics look more for high efficiency when
constructing DCE.

That is why the article of Street et al. is very interesting. Because it let
you construct optimal or near optimal choice experiments. As it is written
on page 463 of the mentioned article: "An optimal design has an efficiency
of 100%. A design is nearly optimal if the efficiency is high but there is
no formal definition of this phrase".

So, as we are looking for high efficient choice design, this method looks
very interesting and suitable.

Nadhem



-----Message d'origine-----
De : limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] De la part de John Rose
Envoyé : mardi 30 mai 2006 00:17
À : Limdep and Nlogit Mailing List
Objet : RE: [Limdep Nlogit List] Choice Model Design Software

Dear Ron and Nadhem

I suggest that in setting up experimental designs that you think carefully
before using the methods spelt out in the Street et al. paper. 
Orthogonality is not a desirable principle in generating designs for
discrete choice models. Orthogonality is designed to produce zero
multicolinearity which is fine in linear models but an unknown in discrete
choice models such as the MNL, where it is the differences in the attributes
that count and not the correlations of the attributes themselves.

I believe that you will find that the second article mentioned will be of
far more use in constructing experimental designs for discrete choice
models.

John Rose

-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of Nadhem Mtimet
Sent: Monday, 29 May 2006 8:49 PM
To: 'Limdep and Nlogit Mailing List'
Subject: RE: [Limdep Nlogit List] Choice Model Design Software

Dear Ron,

Yes it is true that Warren Kuhfeld has developed macros in SAS letting you
getting efficient or near-optimal experimental designs.

As you said the major problem of Warren's method is that it needs few SAS
modules which are very expensive for many persons and institutions to
afford. In that case you will need to buy, SAS Base + SAS STAT+ SAS IML +
SAS QC modules.

Another possibility (very cheap and easy to implement) is to use the method
developed by Street, Burgess and Louviere:

Street, D., Burgess, L., & Louviere, J. (2005). Quick and easy choice sets:
constructing optimal and nearly optimal stated choice experiments.
International Journal of Research in Marketing, 22, 459-470.


This paper is very interesting and let you know how to construct an optimal
or near optimal choice design. You should choose the appropriate orthogonal
plan. This plan will be your first alternative in the choice set. To get the
remaining alternatives you have to "choose some systematic set of levels
changing change to get from the profiles in the first option in the choice
set to the profiles in the second option in the choice set" and so on.... .

I also recommend you to read the paper of Huber and Zwerina (1996), because
it introduces you the theme of choice efficiency and orthogonality:

Huber, J., & Zwerina, K. (1996). The importance of utility balance in
efficient choice designs. Journal of Marketing Research, 33, 307-317.

To get orthogonal plans you can use the SPSS with orthoplan (data /
orthogonal design / generate) or you can use tables of orthogonal arrays
form web pages like Sloane's web page:
http://www.research.att.com/~njas/oadir/

Or  Warren Kuhfeld's orthogonal array web page:
http://support.sas.com/techsup/technote/ts723.html


I hope that this information could help you to construct your choice
experiment.

If you have any problem, please fill free to ask me.

Best regards.

Nadhem Mtimet






-----Message d'origine-----
De : limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] De la part de infocorp
Envoyé : samedi 27 mai 2006 13:23
À : limdep at limdep.itls.usyd.edu.au
Objet : [Limdep Nlogit List] Choice Model Design Software

Dear Limdep and Nlogit List Members,

What software is available for the production of experimental
experimental designs for choice tasks?

I am aware of Kuhfeld's %ChoicEff and associated macros in SAS, however,
they require several SAS Suites which attract rather high annual rental
charges.

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


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