Hello,
In a mediation model where A is the independent variable, B the dependent variable and M the mediator, I observe the following:
- when the difference between M and A is positive (based on average value for the observed group) the mediation effect is partial and positive.
- when the difference between M and A is negative (based on average value for the observed group) the mediation effect is partial and negative.
Can I integrate the difference and use it as moderator in my model and if all indicators are OK (taking care of multicollinearity issue for example), it's OK?
Thank you,
Xavier