[R] choice-based conjoint

Felix Eggers felix_eggers at gmx.de
Thu Apr 1 13:38:08 CEST 2004

Thank you for your answers and suggestioins!

I just read a tutorial on the use of SAS with Discrete Choice Models by
Warren Kuhfeld ("Marketing Research Methods in SAS", 2004, free for
download). He suggests using the Cox PH model. It necessitates reformatting
the data so that the alternatives chosen have a constant survival time of 1
and the other alternatives in the choice set are marked as censored. Thus,
the varying choice sets can be considered in the estimation.

The translation to R seems straightforward, so I will try it out. I might
come back to this list, if I face problems (as I haven't used R before)...


Am 31.03.2004 15:31 Uhr schrieb "stephen.kay at adelphigroup.com" unter
<stephen.kay at adelphigroup.com>:

> The multinomial model can be used- it's the standard method used, although
> there are now more sophisticated ones - see Kenneth Train's website at
> Berkley for the Mixed Logit. I too am new to R so can't really comment on
> how to implement it in this package. If you want to run more sophisticated
> models (e.g. Mixed Logit) and you are an academic (non-commercial) you could
> download a free version of Ox (a matrix programming language very similar to
> GAUSS) and do a search for Mixed Logit - some one wrote a free routine that
> converts Ken Train's GAUSS code to Ox. Afraid I don't have time to look up
> the website addresses.
> Hope this helps,
> Stephen
> Stephen Kay
> Head of Statistics
> Adelphi Group Products
> www.adelphigroup.com
> -----Original Message-----
> From: Felix Eggers [mailto:felix_eggers at gmx.de]
> Sent: 31 March 2004 13:50
> To: r-help at stat.math.ethz.ch
> Subject: [R] choice-based conjoint
> Hello everyone,
> I am new to this list and the R-Project, so I hope my question is not
> trivial or has been answered before. I searched the FAQs and the mailing
> list archives and I could not find anything about Conjoint Analysis. I am
> especially interested in Choice-based Conjoint, resp. discrete choice
> models. 
> Is there a function / module that handles this issue? Or can the multinomial
> logit model be used? I would doubt the latter, for in a Choice-based
> Conjoint analysis the choice has to be conditioned to the alternatives in
> the choice set which are not equal for the respondents.
> Any help will be much appreciated! Thank you!
> Felix
>   [[alternative HTML version deleted]]

More information about the R-help mailing list