You are here

Inclusion of Control Variables in Multiple Mediation Analysis - OSMASEM

2 posts / 0 new
Last post
mjus's picture
Joined: 04/17/2021 - 20:08
Inclusion of Control Variables in Multiple Mediation Analysis - OSMASEM
Image icon Research Model.png37.67 KB

Dear Prof. Cheung & colleagues,

first of all I would like to thank all active members of the forum and of course especially Prof. Cheung for the numerous help and forum entries. The many practical examples and many resources here on the forum are a wonderful source of knowledge and facilitate one's own research immensely.

I am currently working on a project in which I am investigating the mediating influence of several variables using OSMASEM. In the second step of my analysis, I would also like to examine the conditional indirect effect of a moderator variable on some of the mediating relationships in my model. However, as part of the analysis, I have to control for multiple exogenous influences, which presents me with some questions regarding the proper inclusion of control variables in OSMASEM.

On the one hand, I include control variables that I extracted from the correlation tables of my sample (Control Variable 1 in attached Figure) and on the other hand, I have some self-created study-level control variables (e.g., Journal Impact Factor [Control Variable 2 + 3 in Figure]). Now I am wondering how best to incorporate these control variables into my research model.

For the control variables I extracted from the studies' correlation tables (C1), I follow the recommendations of Becker 2005 and include them directly in my model as predictors of my dependent variable.
In contrast, I include the study-level control variables (C2 and 3) as (additional) moderators and specify the paths for whose effect to control via the A matrix.

It would now be important for me to know whether this procedure is correct in principle, or whether there might be a better way which I might have overlooked in order to control for the influence of variables in a research model?

Attached you can find my research model.

Thanks for your time and patience.

Mike Cheung's picture
Joined: 10/08/2009 - 22:37
Control variables 2 and 3 and

Control variables 2 and 3 and moderator variable 1 are treated the same way in the analysis. But you may need lots of studies if you want to include all of them in the same analysis.

If you want to include control variable 1, you need to have the correlations between it and other variables.