I would like to do a mediation analysis as described by Hayes. I am wondering if I need to test for the same assumptions as in regression.
I would usually exclude outliers and extreme values, test for linearity, independent errors, homoscedasticity, normally distributed errors, and no multicollinearity. I have read, that I don’t need to have normally distributed errors, because Hayes method includes bootstrapping.
I have not found anything in the literature on and how to test for the other assumptions. Do I need to test them? If yes, for which path of the mediation do I need to test them? X --> Y or M --> Y? Or both?
Thank you!
The assumptions of the previous mediation is the correlation of variables, and the justification of the model you want to obtain, because mostly you get partial mediation, which requires its theoretical explanation or justify such findings in the discussion, it is not so required to meet the criteria you mention if you use boottrap, this can also be done through free programs available as JAMOVI and JASP, because even in mediating SEM models may be more suitable based on the theoretical model raised.
Assumptions for mediation analysis are same as for regression analysis.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression