This three day course will study the specification, estimation, and application of discrete choice models. We will examine theoretical background and practical application of up to date and frontier techniques in the analysis of microeconometric models for discrete data. Actual studies will be presented. Participants will also apply the techniques using prepared data. Hands on applications will be developed using NLOGIT (developed by the course instructor), one of the world’s most widely used software packages for discrete choice modeling.
The overall nature of the course will be an introduction to discrete choice modeling, with a focus on how to fit, interpret, and use models. Theoretical background will include econometric underpinnings of the models and foundations for various computations, but will not include any derivation, proofs, or establishment of, e.g., asymptotic properties of estimators. The proposed is meant to provide practitioners with the econometric instruction needed to use, understand, and interpret some fairly advanced, yet widely used nonlinear techniques.
William H. Greene is Professor of Economics and Entertainment and Media Faculty Fellow, Department of Economics, Stern School of Business, New York University. More information about ProfessorGreene can be found on his webpage.
The course will include the following topics:
Econometric Methodology:
Estimation, inference and prediction with econometric models; Continuous and discrete choice models; Basics of discrete choice modeling; Binary choice modeling; Models for multinomial choice.
Model Specification:
Random utility models; Utility function specification; Multinomial logit; Nested logit models; Heteroscedastic extreme value; Mixed logit; Finite mixture (latent class); Multinomial probit.
Estimation:
Maximum likelihood estimation; Simulation based estimation; Markov-Chain Monte Carlo estimation techniques; Bayesian and classical estimation procedures; Two step estimation.
Inference:
Estimation procedures; Hypothesis testing; Fit measures and test statistics; Specification testing – the IIA assumption
Data Issues:
Individual and aggregate data: Ordered and unordered choice data; Weighting and choice based sampling; Stated preference and revealed preference data; Merging SP and RP data sets; Heterogeneity; Heteroscedasticity; Panel data; Scaling; Ranks;
Analysis:
Prediction and simulation; Analysis of market scenarios; Specification analysis; Marginal effects and elasticities; Measuring fit in a discrete choice model
Applications:
NLOGIT Software: Examples and case studies from transport, economics, marketing, telecommunications; Hands on applications
Lecture notes, data and papers from past courses can be found in the Resources