Hi, in the attached file, there is moderation testing through PROCESS, IBM, and AMOS analyses. hope the interaction effect (moderation) help you to figure out how to report your result. If you you need further claification, ready to answer.

Hi Zara, it may well be that X affects Y at several levels of M. The point, however, is that these effects will differ in degrees across the M-variable. Actually, the product term signifies the degree of the change of the slope of X when M increases at one unit (or SD, if M is standardized).

If you have not already done it, create a graphic of the interaction effect. This will give you an basis for an interpretation of the nature of the interaction effect.

Best,

Holger

Hello all - thanks ever so much for the quick responses. I'll respond to each of them here to avoid creating a massive thread!

Muayyad - Thanks for attaching your paper - that helps a lot. It's especially interesting to see how you report the insignificant findings, as only one of my interactions produced significant results (out of 8).

Holger - That makes a lot of sense, and has since been suggested to me by my supervisor. She brought a similar paper to my attention (attached) and I think these findings may be similar to mine. I'm going to work on plotting the simple slopes to see how it looks. Thanks again!

Rohit - Thanks for your response. I've been told before to take p-values with a pinch of salt, and generally when I report standard multiple regressions I rely on the CI's to signal significance. I'll take a look at these papers - thanks for attaching them.

Rohit's response leads on to another question actually - for one model, the interaction effect was not significant (in terms of the p value), but X still predicts Y significantly when M is high (but not medium or low). This seems interesting to me and feels as though it still suggests some level of moderation (keep in mind, I am an amateur!)- but from reading countless textbooks on moderation, it looks as though (without a significant interaction effect) I can't infer anything much from that.

So I guess my question is, even if the interaction effect isn't significant - are the conditional effects still of interest?

Thanks again for all your responses so far, and sorry for piling another question on top of it!

Hi Zara,

with regard to your question: Yes it may well be that X has a significant effect on Y when M=1 but not when M=0. Nonetheless the difference between the two slopes (the interaction/moderation/productterm) is not significant. B_X|M=0 may be slightly below the magical significance level, and B_X | M=1 may be slightly above. 

As Rohit correctly said, this is a typical consequence of relying only on p-values. 

Best,

Holger