SEM vs Mediation Analysis
Hi,
I am working on a fairly simple mediation analysis (1 mediator, 1 outcome, 1 predictor, and multiple confounding variables). I was curious if I should use the Mediation Analysis feature in JASP or try to use the SEM process. I keep reading that the benefit of using SEM is to include bootstrapping, but the mediation analysis in JASP allows for bootstrapping. Does the mediation analysis in JASP use Baron and Kenny's approach or the Hayes method? I just want to make sure I am writing my results section properly and describing the process accurately.
Thanks in advance.
Comments
The mediation code is based on lavaan (so SEM), see the help file. So if you build your mediation model in SEM you should get the same result as when you use the mediation analysis. Under "options" in the mediation analysis there is a tick mark for "Show lavaan syntax".
Cheers,
E.J.
Thanks so much EJ. I'm not as familiar with SEM as using the mediation analysis function, so I just wanted to know if I'd get the same results, and if I should discuss the data analysis as Baron and Kenny or Hayes' approach? Or does it even matter? I can't find any articles that just use the mediation analysis, almost all use SEM.
Are you saying I could just copy the Lavaan text from the Mediation analysis and transfer that into my SEM?
It's not using Baron & Kenny. I don't know what the Hayes approach is--what the approach used is is what I think to be the MacKinnon approach, i.e. you multiply the a and b paths and test for significance (the product approach). That's the standard nowadays (and has been for 15+ years), and that's also the approach that Hayes uses (e.g. in his PROCESS macros and his book).
Yes you could copy paste lavaan text from mediation and put that into SEM.
I'm getting an error when trying to copy the Lavaan text over. "Only a subset of population values given. Please specify all population values or none."
I'm not sure what that means. It works in regular SEM, but not in PLS SEM. Why is that?
PLS SEM isn't the same thing and doesn't use lavaan (it's got different syntax). If you want to use PLS SEM (which I'm not sure why you would) then look up syntax for cSEM library in R, that's the library that JASP uses.