I would recommend you to make an analysis of structural equations (multiple regression and path analysis) this model is more robust, the multiple regression is different considering the pearson correlation that it refers (regression infers directed causality unlike the relation), another recommendable analysis considering its multicollinearity would be network analysis, this nonlinear multivariate model is based on independent partial correlations that are not affected by the multicollinearity and they allow you to estimate in a dynamic way the most influential causal relation (similar to the multiple regression). As far as one of its variables has no effect it will not have any significant indirect effect either and they are different methodologies that must be considered according to different specific objectives.

Hello Sidrah,

so your model seems to be that four predictors (yrs. of experience, self-doubt, growth, depletion) have an effect on wellbeing? Why do you bring in mediation? Where's the mediator?

What I wonder ist, whether years of experience has an indirect effect effect on the other three predictors--in this case this would indeed be a mediation model and it would explain the non-sign. direct effect.

Best,

Holger

Holger Steinmetz Hi, thank you for your response.

I was trying to include growth and depletion as moderators of the link between self-doubt and well-being to explain the multicollinearity found between growth, depletion and self-doubt. There is previous research to suggest self-doubt has less impact on therapists well-being when they experience growth and more when they experience depletion. My intention was to use process macro to explore whether there is a link.

I have added a summary of the statistics if this helps

Again, thank you so much

Hi Sidrah,

actually your correlation between depletion / growth and wellbeing are so large that I have rather the impression that they are the same (isn't depletion a kind of exhaustion?)

At any rate: With such a high correlation, you have to chance to test moderation, as the moderator and the dependent variable are practically the same.

Best,

--Holger