Myths About Measurement, Causality, and Structural Equation Models
Ken Bollen is a professor in the Departments of Psychology & Neuroscience and Sociology at UNC at Chapel Hill, as well as head of the Methods Core and a Fellow at the Carolina Population Center at UNC. From 2000 to 2010, he was the Director of the H.W. Odum Institute for Research in Social Science at UNC. His current research is on the development of statistical methodology for the social and behavioral sciences.
Abstract: Structural equation models (SEMs) are a common tool to analyze social, behavioral, marketing, and health data. The popularity of SEMs is partially due to their ability to incorporate latent and observed variables, to analyze mediating relations, and to control for measurement error. Their generality as statistical models also contributes to their frequent application. Accompanying their widespread use are widespread misunderstandings about their properties. This talk will review a number of myths about causality and measurement and their role in SEMs. The goal is to clarify the power and the limits of SEMs by sweeping aside the myths.