Presentation at the 2006 Annual Meeting of the Association for Psychological Science (APS)

Mediation Analyses in Longitudinal Studies

Patrick E. Shrout

Department of Psychology

New York University
Mediation analysis provides tools for understanding processes associated with experimental and nonexperimental relations.  In this talk I review conceptual and statistical issues that arise applying mediation concepts to longitudinal studies, including the specification of competing mediation models and choosing among alternative statistical approaches. Examples are from prevention trials in children.

Copy of Powerpoint slides

Click HERE for an example of SAS syntax that reads and analyzes an SPSS generated
data set in .por format.  The lines in this data set are time points for each child.  The
example data can be obtained by clicking HERE.

The example shown above is simulated to resemble data that we are currently
analyzing.  They are not real data, so do not get excited by your findings.  If you
want to see the SAS program that simulated the data, click HERE.

The simulated example does not have any missing data, but you can eliminate
followup points for different children and verify that the generalized linear model
program still recovers the answer.  The data can be missing at random, which includes
the likely possibility that more data will be missing for later time points.