American Statistical Association
The usual approach to mediation in the social and behavioral sciences uses a structural equation model, decomposing the total effect into direct and indirect effects. Using potential outcomes notation and the approach to causal inference that has developed in statistics over the past 35 years, it is easy to see that the parameters of structural equation models should not be interpreted as effects, unless special conditions, typically unlikely to be met in practice, hold. After giving a brief introduction to causal inference, the claims above are shown and then some possible alternatives to conventional structural equation modeling are offered and discussed.
Michael Sobel, Ph.D. is a Professor in the Department of Sociology at Columbia University. He received a Ph.D. in Sociology from the University of Wisconsin Madison. Dr. Sobelís statistical research has focused on causal inference, models for categorical data, and structural equation models. His substantive research has been primarily in the area of social stratification. Current interests include neighborhood effects and applications of finance.
|Date:||Tuesday, November 30, 2010|
|Time:||3:00 - 4:00 P.M.|
New York State Psychiatric Institute
1051 Riverside Drive
6th Floor Multipurpose Room (6602)
New York, New York