Leeds University
CMC

Software

The Choice Modelling Centre (CMC) at the University of Leeds has developed flexible estimation code for choice models in R. The code uses the complete opposite of a black-box approach, i.e. the user sees every step in the coding of a log-likelihood function. We believe this to be essential in ensuring a greater understanding by users of the very powerful models they have at their disposal.

CMC has developed code for all different types of choice models, including but not limited to Multinomial Logit, Nested Logit, Cross-Nested Logit, Mixed Logit, Latent Class, Mixed GEV and hybrid choice models. We have also produced code for the estimation of Multiple Discrete Continuous Extreme value models, mixed logit models allowing for both inter and intra-respondent heterogeneity, as well as code using alternative estimation approaches, including Bayesian techniques and the EM algorithm. The code has been tested extensively and is relatively robust, although the reliance on numerical derivatives will of course lead to occasional issues. We accept no liability for any use of the code.

The code we make available to users on our website covers simple implementations of Multinomial, Nested and Mixed Logit on a mode choice data set with four alternatives (car, air, rail and high speed rail). No detailed documentation is provided nor is there any support available. For users interested in learning more about our code, CMC runs two annual courses that combine theoretical lectures in choice modelling with an in-depth introduction to the R estimation code. 

A basic understanding of R is required to use the code. Background reading is for example available at https://www.r-project.org/about.html and http://www.statmethods.net/

We ask users of the code to cite it as follows:

CMC (2017), CMC choice modelling code for R, Choice Modelling Centre, University of Leeds, www.cmc.leeds.ac.uk

Pdf with CMC code introduction.
The code can be downloaded here

Examples of CMC publications that have used this code:

Calastri, C., Hess, S., Daly, A.J., Maness, M. Kowald, K., Axhausen , K.W. (2017), Modelling contact mode and frequency of interactions with social network members using the multiple discrete-continuous extreme value model, Transportation Research Part C, 76, pp. 16-34.

Lu, H., Hess, S., Daly, A.J. & Rohr, C. (2017), Measuring the impact of alcohol multi-buy promotions on consumers’ purchase behaviour, Journal of Choice Modelling, accepted for publication, May 2016.