Workshop 7: Causal Mediation Analysis in Multi-site Trials

The 7th Workshop on Quantitative Methods in Education, Health and the Social Sciences (QMEHSS) will be on Friday April 18th from 11:00-12:30 and will be co-presented by Dr. Guanglei Hong and Xu Qin . The seminar will be held in the NORC conference room 344. NORC is located at 1155 E. 60th Street.

Causal Mediation Analysis in Multi-site Trials

Xu Qin and Guanglei Hong

Department of Comparative Human Development, University of Chicago

In analyses of data from multisite randomized trials, researchers obtain evidence with regard to not only the overall average treatment effect but also the variation in the treatment effect across sites. Further research is then required, though rarely done in the past, for investigating the prevalent causal mechanism and its heterogeneity across sites. We use the national Head Start Impact Study as a motivating example. The causal effects of interest include the indirect effect of assignment to Head Start programs on child vocabulary learning mediated by a program-induced increase in parent reading to child and the average direct effect of Head Start programs. Multilevel path analysis and structural equation modeling (SEM), among their many other constraints, do not have a single parameter corresponding to the average indirect effect. Extending the ratio-of-mediator-probability weighting (RMPW) approach to causal mediation analysis in multisite trials, we estimate not only the average indirect and direct effects but also the between-site variance and covariance of these effects. We explain the theoretical rationale of the RMPW approach and assess the performance of our estimation strategy through simulations.