Must the moderator variable correlate significantly with the dependent variable, if I am to do a moderation analysis (multiple regression)?
I have a predictor (X) and an outcome (Y). They correlate significantly. My hypothesis is that a moderator (W) moderates the relationship. Multiple regression analasys shows that this is true. M is a significant buffer, weakening the correlation between X and Y. BUT W does not correlate with X or Y ... I see that W correlates with Y in most moderation analyses - but I cannot find anyone who explains, if this is required, to do the analysis in the first place!?