Could you recommend an easy to use or comprehensive conjoint analysis package for R?
I've never used R for conjoint analysis, but here are a couple of things I found when I hunted around.
- Aizaki and Nishimura (2008) have an article "Design and Analysis of Choice Experiments Using R: A brief introduction" (Free PDF available here).
Perhaps check out the following packages:
mlogit is the best R package I've found for modelling discrete choice data. It supports the basic multinomial logit, as well as more advanced models such as multinomial probit and mixed logit. The package also includes specification tests to choose between different models.
You may want to use faisalconjoint package in R, it is tested with many published and research data, it works perfectly, one on important thing its works without design restriction and rank procedure. It works in all condition and provide accurate estimates.
The best in my opinion for R is a conjoint package from CRAN: http://cran.r-project.org/web/packages/conjoint/index.html
If you are looking for models other than logit,
- you can use 'survival' package to build conditional multinomial logit model.
- you can use 'bayesm' package to build hierarchical bayesian(HB) model. Sawtoothsoftware asked the guy who created this package to help them build HB model in their software.
Faisal Conjoint Model (FCM) is an integrated model of conjoint analysis and random utility models, developed by Faisal Afzal Sid- diqui, Ghulam Hussain, and Mudassir Uddin in 2012. Its algorithm was written in R statistical language and available in R . Its design is independent of design structure. It could be used for any research design i.e. full prole, orthogonal, factorial, supersaturated etc. Another important point about FCM is rank procedure. It works for every kind of ranks i.e. unique ranks, percentage ranks, tight ranks, missing ranks etc. It has been tested for many published data. Most of the times, FCM results are same with same magnitudes, often the rank
There is a library 'Conjoint' with many features and sample to find utilities. For a quick preview check the link. This will help you get started.
"survival" (clogit) for multinomial logit (MNL) model.
"mlogit" for a wide range of models (MNL, nested logit, heteroscedastic logit, mixed logit (MXL) also known as random parameters logit, ...).
In the same spirit you should take a look at "Rchoice" (file:///C:/Users/kruci/Downloads/v74i10.pdf).
"bayesm" for bayesian version of MNL/MXL - However if you are interested in bayesian approach I would strongly recommand the great "RSGHB" package.
"gmnl" for the generalised MNL model.
"flexmix" for latent class logit (LCL) model.
More generally it is important to keep in mind that choice models are a special case of multilevel (or hierarchical) models (you have choices nested within participants themselves nested within higher units: supermarkets, countries, etc.) - So everything that can be used for multilevel modelling (e.g., the great "lme4" package) and that can also accommodate the discrete nature of the choice variable would do the job. For example, you could use "lme4" if the choices are binary (Do you want this product? Yes/No) or made between 2 options (Which product do you want? A/B).
With Stata, you have many commands useful for choice modelling:
clogit for MNL
mixlogit for MXL
clogithet for heteroscedastic MNL
lclogit for latent class logit
gmnl for generalised MNL
Many of these commands have been developped/refined by Arne HOLE (Great job!) http://www.stata.com/meeting/uk13/abstracts/materials/uk13_hole.pdf
Choice modellers also use other software: nlogit (developped by W. Greene) biogeme (Thanks to M. Bierlaire) - Great tool but can only be used for choices modelling I've heard about LatentGOLD but not sure ...
For those who want to use MATLAB, You got to take a look at:
Mikołaj Czajkowski webiste (http://czaj.org/research/estimation-packages/dce)
Kenneth TRAIN website (https://eml.berkeley.edu/~train/software.html) - Actually most of the choice functions come from Kenneth TRAIN's work
Finally, for those who are willing to invest a significant amount of time in the coding of choice models, Chandra BHAT website is amazing (http://www.caee.utexas.edu/prof/bhat/FULL_CODES.htm)
Many thanks to all these great researchers (Train, Bhat, Bierlaire, Hole, Croissant, Czajkowski, etc) who made this possible!