Spring 2008 Seminar Series

DATE: March 20, 2008
LOCATION: E40-298
TIME: 4:15pm
Reception immediately following in the ORC ConferenceRoom, E40-106

SPEAKER:
John Liechty

TITLE
Approaches to Heterogeneity: Adventures with Mixture Models

ABSTRACT
One of the main tasks in Marketing is developing models that provide an understanding of empirical demand functions. Typically empirical demand functions are built up from models that accommodate individual level demand; in practice these resulting models can be calibrated from survey based or experimental based studies such as a conjoint study. Often these demand functions are modeled via mixture models. In this talk, I discuss two new approaches to mixture modeling. The first approach is very similar to the standard finite mixture model, with the expectation that information for the mixture components are aggregated at the margin. This approach results in a surprisingly powerful model that is able to uncover complex structure with a relatively low level of information, when compared with the standard finite mixture model. The main challenge with this model is with regards to calibration and model choice. I will discuss a Reversible Jump Markov Chain Monte Carlo algorithm which allows for an efficient and robust approach to calibration. The second approach provides a mixture model that is both flexible and has clearly identified segments or groups. This approach was originally motivated by a problem from modeling eye movements. Both approaches will be illustrated with simulated data and in the case of the first model, data from a field study.