Data Analytics Corp.
2011-Jun-21 14:45 UTC
[R] Estimating choice models at the individual level
Hi, I have a discrete choice model to estimate for a client that I originally planned to estimate as an aggregate model using a clogit routine. Now the client is asking for results for many segments of the respondents which would mean, if I stayed with my original plan, I would have to estimate a large number of models. I could certainly do this, but I'm thinking that it would be better to estimate a Hierarchical Bayes model to get coefficients for each individual respondent. This way, I could pull out the people I need for a particular segment and use just those coefficients. Sawtooth's program for MaxDiff can do this. Is there any R package to do an HB estimation for a discrete choice (logit) model? Thanks, Walt ________________________ Walter R. Paczkowski, Ph.D. Data Analytics Corp. 44 Hamilton Lane Plainsboro, NJ 08536 ________________________ (V) 609-936-8999 (F) 609-936-3733 walt at dataanalyticscorp.com www.dataanalyticscorp.com
The "bayesm" package does HB models for discrete choice. In practice, a hurdle is likely to be translating your data (if it's fielded in Sawtooth Software) to the correct format for bayesm (which uses a list format instead of a matrix, and difference coding instead of dummy coding). (BTW, note another R list: R-SIG-DCM, which you might consider joining if you're not already a member. https://stat.ethz.ch/mailman/listinfo/r-sig-dcm ) Good luck, -- Chris -----Original Message----- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Data Analytics Corp. Sent: Tuesday, June 21, 2011 7:45 AM To: r-help at r-project.org Subject: [R] Estimating choice models at the individual level Hi, I have a discrete choice model to estimate for a client that I originally planned to estimate as an aggregate model using a clogit routine. Now the client is asking for results for many segments of the respondents which would mean, if I stayed with my original plan, I would have to estimate a large number of models. I could certainly do this, but I'm thinking that it would be better to estimate a Hierarchical Bayes model to get coefficients for each individual respondent. This way, I could pull out the people I need for a particular segment and use just those coefficients. Sawtooth's program for MaxDiff can do this. Is there any R package to do an HB estimation for a discrete choice (logit) model? Thanks, Walt ________________________ Walter R. Paczkowski, Ph.D. Data Analytics Corp. 44 Hamilton Lane Plainsboro, NJ 08536 ________________________ (V) 609-936-8999 (F) 609-936-3733 walt at dataanalyticscorp.com www.dataanalyticscorp.com ______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.