How do I handle four latent variables in SEM? Out of the four I have one to be a mediator, one to be a moderator, one as a dependent variable, and lastly one to be an independent variable. It's a relational study. How do I handle all the four in one model as a beginner in SEM? Thank you.
Sylvester Kwesi Dadzie Examining mediation with latent variables is fairly straightforward (see, e.g., MacKinnon, 2008). Moderation (studying latent interaction effects) is more tricky. There is a framework called "latent moderated structural equation modeling" (LMS; Klein & Moosbrugger, 2000) that is implemented in Mplus that allows you to specify latent interaction terms in a straightforward way and that appears to work well in many cases.
A general suggestion would be to begin by examining the measurement structure first by running a confirmatory factor analysis with all four factors and allowing the factors to freely correlate (saturated structural model). If this 4-factor correlated CFA model fits well and provides reasonable/plausible results, you could try the mediation (latent path) model next, which is relatively straightforward to specify in most SEM programs. The latent moderation (and/or combined mediation/moderation) model is most complicated and could be examined last.
Klein, A., & Moosbrugger, H. (2000). Maximum likelihood estimation of latent interaction effects with the LMS method. Psychometrika, 65(4), 457–474. https://doi.org/10.1007/BF02296338
MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. Taylor & Francis Group/Lawrence Erlbaum Associates.
Thanks a lot Christian Geiser for sharing this book
"Introduction to statistical mediation analysis. Taylor & Francis Group/Lawrence Erlbaum Associates"
I am wondering whether this (suggested treatment) applies in both Co-SEM and PLS-SEM?
Thanks
Salah Ahmed To the best of my knowledge, MacKinnon's (2008) book on mediation does not cover PLS.
- Badges
- Science method