Moderation - Which Variable?

[LisaB]

Hi, I am running moderation analysis using regression in SPSS. I have read that when you have a significant moderation interaction either of the interacting IVs can be said to be the moderator. I am wondering if this is correct, or whether there is and indicator (other than theory) about which one should be moderating the relationship? I ask this because sometimes when I run simple slope analysis I find that if I run simple slopes for conditions of the moderating variable on the IV/DV the slopes are not significant, but if I run the other IV as the moderator I do get significant slope(s) for the IV/DV relationship at different conditions of the moderator. Does this indicate that the moderator should be the IV that provides significant slope gradient(s)?

Any advice would be appreciated.

Lisa

 

[Ktau]

As you have read, either variable in an interaction can be considered to moderate the relationship between the other variable and the outcome. Which variable is assigned the label of moderator in a relationship depends on the research question/hypothesis. Moderation does not imply a causal relationship in any way, only that the relationship between two variables differs as a function of a third variable.

As for your simple slope analyses, consider the attached image.

These are the same data graphed differently. The slopes change depending on which variable I make the moderator (series). In a similar manner, simple slope analyses often differ depending on which variable you make the moderator because the slopes being examined are fundamentally different. The underlying moderated relationship remains the same regardless.

I have seen people switch labels based on personal preference for one simple slope analysis or another, but simple slope analyses are not an indication of which variable "should" be labeled the moderator.

 

[LisaB]

Thanks so much for your input. I feel reassured now that I'm not missing something.

Best wishes
Lisa

 

[noetsi]

To me the term moderator as it is being used here is confusing. It seems what is being talked about is actually interaction. Interaction occurs when the impact of the IV on the DV (the slope) varies at different levels of another IV. A moderator is less well defined, as far as I can tell, but commonly involves one IV masking the impact of the IV on a DV (that is making it small or reversing the sign).

A moderator in the general usage is not an interaction effect.