I have a question about using multiple regression in Process. I have a scale with four subscales. I would like to investigate the mediation effects of these four subscales on the link between my independent and my dependent variables. When I do 4 separate mediation analyses, two of my medications reach a significant level. However, when I do one multiple mediation analysis, I got no significant mediation. The Correlations between the subscales ranges between .3 to .5.
I am not sure whether this level of correlation among the variables is strong enough to consider collinearity a problem and which approach is better: separate mediation analyses or 1 multiple mediation analysis?
Appreciate for your help,
Thanks Alvin for your response. They are correlated as they are the sub-scales of one measure.
Dear Somayyeh
I hope I got you right. You intend to conduct a multivariate analysis. To be frank and conservative, in order to avoid Type 1 error, even though you do separate mediation analyses, please employ Bonferroni Correction: p-value / No. of analysis In your case, since you have four separate mediation analyses, supposing the p-value = .05. After Bonferroni Correction, .05 / 4 = .015. You actual p-value will be deflated to .015.
Nevertheless, when you do to multivariate mediation analysis, you will NOT have to bother with Bonferroni Correction.
I understand it is a dilemma. I really hope my words will help otherwise your analysis would be subject to Type 1 error.
Happy new year!
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude