Monte Carlo Simulation and Resampling Methods for Social Scientists (Chapel Hill, NC)


As data sets become larger and more diverse, standard substantive theories and/or statistical methods may not be appropriate for many research applications. The question is: how can you know when they are and when they are not? Monte Carlo simulation and resampling methods allow researchers to explore and often relax both theoretical and statistical assumptions, and thus, to better exploit the data they have.

These methods are rigorous, robust, and widely used; thereby making them an increasingly essential addition to any empirical social scientist's toolbox. These methods achieve their benefits by allowing researchers to use their computers and data as "experimental laboratories" for data analysis. They foster the development of deeper intuition and understanding through logic and visualization without requiring a strong foundation in mathematics, making these methods accessible to virtually all empirically oriented social scientists.

This class will examine both the similarities and differences between Monte Carlo simulations one the one hand and resampling methods on the other. We will begin by exploring (through simulation) that oft-repeated phrase in statistics classes – "in repeated samples" – as well as the fundamental role of theory and randomness in social science. We will then turn to a number of applied and practical examples. We will explore simulations for liner and general linear models, clustered and pooled data, and both cross-sectional and dynamic data. Topics in resampling will include bootstrapping, permutation and randomization testing, posterior sampling, and cross-validation. Participants will also have the opportunity to add topics to the list of those covered in the course, which might include MCMC, other simulation methods, and/or simulation based tests of competing models. Students will work through numerous examples in a computer lab, learning to present results numerically and graphically.

Participants should have at lease some basic familiarity with standard multiple regression. Familiarity with R will be helpful, but it is certainly not required. The course will follow the content of the book Monte Carlo Simulation and Resampling Methods for Social Science by Thomas M. Carsey and Jeffrey J. Harden (Sage, forthcoming).

Fee: Members = $1500; Non-members = $3000

Tags: Monte Carlo, Simulation, Resampling,

Course Sections