This workshop introduces hierarchical/multilevel regression-style linear models (and a few basic nonlinear forms) in a manner accessible to social and behavioral scientists. These models account for levels of aggregation that are typical in social science data in which individuals are nested in groups, and possibly multiple groups. Since these specifications are inherently Bayesian in nature this workshop will also introduce the basic principles of Bayesian statistics to students in the social and behavioral sciences without requiring an extensive background in mathematical statistics. Most of the examples will be drawn from sociology, political science, economics, marketing, psychology, public policy, and anthropology.
The prerequisites for this workshop are a linear regression course and knowledge of matrix algebra. The emphasis will be on applying the principles to actual data-analytic problems of interest to participants rather than through textbook examples. The workshop will make extensive use of software that is in the public domain, yet is high in quality (including R and JAGS), please make sure to bring a personal laptop with both programs installed.
Workshop Outline & Reading List
Thursday, January 9
Advantages of Multilevel Models
Features of Multilevel Models
Linear Model Illustration
Comparison with Variable Contrasts
Partial Pooling Estimates with No Explanatory Variables
Contrasting Pooling Approaches
Presenting Results from Multilevel Models
Simple Illustration of Bayesian Inference
Specifications with the lmer() Function
A Bayesian Take on Hierarchical Models
Panel Data as Group Membership
Varying Intercept Logit Multilevel Model
Bayesian Multinomial Specifications for Employment Status
Random Effects Example for Indomethacin Trials
Nested Classification Factors
Simple Linear Bayesian Specification: Poverty Among the Elderly, Europe
Prior Sensitivity, ANES Data from 2012
Logit Model for Survey Responses in Scotland, Percent Predicted Correctly
Another application: Poisson Model of Military Coups
King, Gary (1986) “How Not to Lie With Statistics: Avoiding Common Mistakes in Quantitative Political Science”. American Journal of Political Science, 30, 666-687, 1986.
Gill, Jeff and Andrew J. Womack (2013) “The Multilevel Model Framework”. In The SAGE Handbook of Multilevel Modeling. Scott, Marc A, Jeffrey S Simonoff and Brian D Marx (eds). London: SAGE Publications Ltd, pp. 3-20. SAGE Research Methods.
Gill, Jeff and Chris Witko (2013) “Bayesian Analytical Methods: A Methodological Prescription for Public Administration”. Journal of Public Administration Research and Theory, 23:2, pp. 457-494.
In 2020, for the first time, the WSSR is collaborating with the Southern Political Science Association and is hosting a series of half- and full-day workshops during their conference in San Juan, Puerto Rico.
Join us between January 8th and January 11th and attend one or more of our workshops: