I am conducting a moderated mediation analysis using Hayes' Process Macro Model 10. The analysis reveals that the indirect effects are only significant for some values of the moderators. The index of moderated mediation is insignificant. What can I say about the effects of the moderator?
In detail, the assumed mediation is only significant when the moderator (product involvement) is low or moderate. Does this mean the manipulated variable acts as an peripheral cue and loses its impact for high involved consumers - so the results show a moderation? Or does the insignificant moderated mediation says the product involvement can not be seen as a moderator?
Thanks!
Much like the product or interaction term in a linear regression analysis quantifies the relation between a predictor and a moderator, the index of moderated mediation quantifies the relationship between the indirect effect and a moderator.
My opinion: apparently you chose the Hayes approach to assess if you have evidence for a moderating effect on the mediation. Therefore, you also chose the index of moderated mediation as an indicator for this hypothesis (apparently, otherwise you would not have calculated it). As Hayes claims, it is more useful to assess this hypothesis with the whole model instead of the pieces. Now, your index of moderated mediation was not significant, i.e. you have no evidence that supports your hypothesis (strictly speaking).
And just like post hoc tests in an ANOVA model, it is not recommended to analyse the pairwise comparisons, if the omnibus test is not significant. Even worse in regression with metric moderators: you do not have fixed groups, but you can pick and choose where you want to test the slopes. Unless the moderating effect is perfectly zero, you can expect that the model will show slopes different from zero at any artificial point of the moderator, so this is not really something to wonder about.
Additionally, keep in mind that the interaction and hence the index of moderated mediation tests, if the slopes change with changes in your moderator(s), but the "post hoc" simple slope analyses test, if the (tested) slopes are signficantly different from zero, which is a different question. Maybe all slopes are different from zero (simple slopes) but not different from each other (interaction/moderation). You may have an edge case, where some may be slightly different from zero and others not. Did you do an a priori power analyses to determine the appropriate sample size? Did you check the effect sizes? Did you asses a priori the smallest effect size of (practical) interst (SESOI) for your study? If not, you cannot say more that you do not have enough evidence to claim a moderated mediation.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS