I am using the PROCESS macro (model 7) to examine a moderated mediation and ran into a new problem.
Bivariate correlations show X is positively correlated with Mediator, but not with Moderator.
Moderator is not correlated with X or mediator.
When I run the moderated mediation, the Interaction is term is significant and the results are consistent with theory. However, X and Moderator now show negative loadings on Mediator (though only Moderator significantly so).
I understand this is a likely a form of suppression. But am not sure what to do. Is it acceptable to publish as is and just interpret the interaction? Does this indicate some flaw in the model? Thanks!