

Confounders in path models 

Message/Author 

Jon Heron posted on Thursday, May 10, 2012  5:50 am



Hi Bengt/Linda I have a question about how to correctly introduce a confounder for one or more paths in an SEM/path model. Say we have a standard mediation model Y on X M; M on X; If we have a confounder C that impacts on the main effect from exposure X to outcome Y, then the additional command Y on C; will allow for this, presumably by assigning C to be an exogenous variable which is correlated with X. In a similar way the line M on C; will adjust the path from X to M. So my question is, how does one adjust the path from M to Y for confounding, either on it's own or in conjunction with the other two described above. I'm assuming that the model Y on X M; M on X; Y on C; M on C; only adjusts two paths, rather than all three. Many thanks, Jon 


You may want to take a look at Section 10 of Muthén, B. (2011). Applications of causally defined direct and indirect effects in mediation analysis using SEM in Mplus. which is on our web site under Papers, Mediation Modeling. Section 10 describes the sensitivity analysis proposed by Imai, inspecting the effect of confounding of the M>Y relationship. This confounding sensitivity analysis is very easy to do in Mplus. 

Jon Heron posted on Thursday, May 10, 2012  1:54 pm



Thanks Bengt I've been carrying that one around for a while, sounds like reading it would help :) 

Back to top 

