David F Layton
Examines various research topics of importance in public policy and management.
TOPIC: DISCRETE CHOICE- This offering of PPM 599 focuses on discrete choice analysis. Discrete choice analysis considers the statistical modeling of choices made by agents (individuals, households, groups, firms, and other entities). The discrete choices might take the form of answering yes/no questions, voting, choosing one house, trip, or commute route out of many, and other related problems such as evaluative ratings or rankings. These choices might be one time decisions or repeated through time. It also relates to more general models where a continuous dependent variable is not observed over its entire range for some reason (e.g. Tobit models). The dimensionality of the choice might be dichotomous or very high dimensional. This branch of econometrics and statistics has become increasingly important in empirical work in many areas of economics and related disciplines. Recent innovations in computational and econometric methods focusing largely on simulation based approaches to estimation have seen even wider application of ever more realistic but more complicated discrete choice models. The goals of this course are applied in that the focus is on working with data, estimating models, and in conducting empirical economic analysis. Specifically, the goals include: learning to write estimation programs; to implement and then analyze the results of simulation estimators for high dimensional discrete choice problems; and to conduct applied analysis of welfare measures and other quantities in the context of choice models. While the models considered here are discrete choice, extensive practice in estimation, testing, and analysis builds important skills that are readily transferable to other classes of statistical problems. Experience indicates that learning to code one’s own programs is an excellent way to learn the methods even if one ultimately uses routines in available software. Grading will be based on a set of coding and estimation projects. The text for the course will be Discrete Choice Methods with Simulation by Kenneth E. Train, Cambridge University Press, 2003. There will be supplemental readings as needed. The prerequisite is graduate level statistics/econometrics at the level of SOC 505/506 or ECON 581/582 or similar. Graduate students from all disciplines welcome.
Student learning goals
General method of instruction
Prerequisite: graduate level applied statistics/econometrics coursework at the level of SOC 505/506 or ECON 581/582 or similar. Please contact the Professor with questions regarding preparation.
Class assignments and grading
Coding statistical/econometric programs for discrete choice analysis