We want to explore a mediating relationship employing SEM with Amos SPSS. Our independent variable is attitude toward wife abuse. mediator is exposure to male marital violence, and outcome variables are women's agency and mobility. There seems a significant mediating effect. However, we want to see if the relation remains significant after holding constant a few control variables: age, education, years married, and income. Can we control these variables during testing the above mediation effect using Structural Equation Modeling with Amos?
You can control any variable in SEM using AMOS. You can see the below paper how to use the control variables in AMOS.
https://idp.springer.com/authorize/casa?redirect_uri=https://link.springer.com/article/10.1007/s10668-020-00951-3&casa_token=YypcA7nLCbAAAAAA:zXPrFOcg9zmOBeYZ9qHJy2c4iIjdtG_1vNGe1KijJm0NzjA72VBaE1GccVC8OctlIxXQONXiuksfR-QkwHk
You can but you must specify well the study objectives, since it is not always recommended to use them.
Hello Rabiul,
I wrote a possibly helpful post in this thread with some references.
https://www.researchgate.net/post/What_is_the_role_of_control_variables_in_SEM#view=557d9d275cd9e3a0a58b45b6
In a nutshell, a control variables serve to eliminate a non-causal link between your IV and our DV (may it be the overall outcome or the mediate) which existence biases your estimated effect. Non-causal links are those created by unmeasured confounders--that is variables that causally affect your IV and your DV. The respective control variable--to be effective for reducing the bias--may be the confounder itself or a variable that mediates the effect of the confounder on the IV or DV. Think in terms of a electric circuit in which you shut off one of the inner relais.
Controlling in an SEM can be done by the following:
a) The control variable correlate with the IV and has an effect on the mediator and/or the DV
b) (Better): The control variables has effects on the IV and the mediator OR the IV and the outcome OR the mediator and the DV. If it is reasonable the control variable may affect all three.
The mentioned post also explains those kinds of variables that must not be controlled (mediators, colliders, and instruments).
In a summary, control variable are part of your theoretical model and should be modelled as such.
HTH
--Holger
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling