Abstract
We represent a linear structural mean model (SMM)approach for analyzing mediation of a randomized baseline intervention's effect on a univariate follow-up outcome. Unlike standard mediation analyses, our approach does not assume that the mediating factor is randomly assigned to individuals (i.e., sequential ignorability). Hence, a comparison of the results of the proposed and standard approaches in with respect to mediation offers a sensitivity analyses of the sequential ignorability assumption. The G-estimation procedure for the proposed SMM represents an extension of the work on direct effects of randomized treatment effects for survival outcomes by Robins and Greenland (1994) (Section 5.0 and Appendix B) and on treatment non-adherence for continuous outcomes by TenHave et al. (2004). Simulations show good estimation and confidence interval performance under unmeasured confounding relative mediation approach. Sensitivity analyses of the sequential ignorability assumption comparing the results of the two approaches are presented in the context of two suicide/depression treatment studies.
Disciplines
Statistical Models
Suggested Citation
TenHave, Thomas R.; Joffe, Marshall; Lynch, Kevin; Brown, Greg; and Maisto, Stephen, "Casual Mediation Analyses with Structural Mean Models" (November 2005). UPenn Biostatistics Working Papers. Working Paper 1.
https://biostats.bepress.com/upennbiostat/art1