Guest Speaker: Dr. Trent Mize / Purdue University

Please join the Computational Social Science Workgroup of the Owens Institute for Behavioral Research and UGA Sociology as we welcome Dr. Trenton D. Mize (Purdue University) for his presentation titled: "The Problem With “All Else Equal”: Generalized Marginal Effects and Better Answers to Real-World Causal Questions"

Abstract: To quantify effects, health and social science researchers typically estimate a regression model and interpret the coefficient, which represents the effect of a change in a variable holding "all else constant." But real-world causes rarely operate in isolation—changing one variable often changes others along with it, and the same cause can play out differently across groups with different characteristics. In this talk, I introduce generalized marginal effects, which extend standard marginal effect approaches to allow new possibilities for matching causal estimands to theoretically motivated questions. For example, generalized marginal effects can capture the impact of multiple variables changing together, disentangle the downstream consequences of a cause, and determine why effects differ across groups. The methods have widespread applications for observational and experimental data, as well as for predictive modeling and machine learning approaches. I illustrate the utility of each application using large, publicly available social science datasets and provide resources for implementing the methods in Stata and R.
 

About Dr. Mize:

I am the dean's associate professor of the College of Liberal Arts in the departments of Sociology and Statistics (by courtesy) at Purdue University. I am also a founder and co-director of the The Methodology Center at Purdue and a founder and co-director of the Kernan Experimental Social Science Lab. I am a quantitative methodologist specializing in categorical data analysis, data visualization, latent variable modeling, and experimental design. I am also a social psychologist and use theories of identity, status, and stereotyping to understand (a) how social categories impact how we view ourselves and how others view and treat us, and (b) the social factors that influence health and well-being. I teach courses and short workshops on applied statistics, quantitative methods, and social psychology.