Search
Program Calendar
Browse By Day
Browse By Time
Browse By Person
Browse By Room
Browse By Unit
Browse By Session Type
Search Tips
Visiting Washington, D.C.
Personal Schedule
Sign In
X (Twitter)
We develop methods for causal mediation analysis in multisite randomized trials under the potential outcomes causal framework. The study is motivated by a re-evaluation of Job Corps programs nationwide. One important advantage of the multisite design is that it becomes possible to reveal not only the prevalent causal mechanism but also possible heterogeneity of mediation mechanisms across the sites. Such information may greatly inform program design and implementation. We develop a weighted method-of-moments procedure that consistently estimates the population average and the between-site variance of the direct and indirect effects. Unlike multilevel path analysis and SEM, the weighting strategy relaxes the assumption of no treatment-by-mediator interaction while greatly simplifies not only the outcome model specification but also the estimation.