The course aims at introducing Bayesian methods for quantitative marketing. The course is structured in two parts. In the first part, students will be provided with a theoretical background in Bayesian statistics and computational methods with an emphasis on the practical advantages of the Bayesian approach, implementation and informative priors. The second part of the course will discuss models commonly encountered in marketing, how to estimate them and how to implement estimation (with R).
Exam: The exam will consist of a project based either on a research article or on a case study where students can choose among: developing a simulation study, developing a real data analysis, comment and extend the paper.
- Introduction: Bayesian approach versus frequentist approach in marketing. The importance of decision-making in marketing.
- Basic concepts of Bayesian statistics: essential topics in Bayesian decision theory, prediction, conditioning and the likelihood principle, assessment of the prior distribution, point and interval estimation, testing and model choice.
- Classical models: Regression and Multivariate Analysis, hierarchical models, mixture models, etc..
- Markov Chain Monte Carlo methods: Gibbs sampler, data augmentation, Metropolis algorithms.
- Latent variable models, Multinomial and Multivariate Discrete Models, Demand Theory for Models of Discrete Choice.
- Hierarchical Models for Heterogeneous Units (heterogeneity of customers’ preferences).
- Simultaneity (demand-supply models, endogeneity, instrumental variables).
- Case studies.
- Bayes and Big Data in Marketing.
P.E. Rossi, G.M. Allenby and R. McCulloch, “Bayesian Statistics and Marketing”, 2005, Wiley.
Other references to journal articles will be provided during the course.