Question regarding moderation analyses....

[grenas]

This community has been supportive and I've searched posts regarding moderation but did not find what I needed. Please be kind to me. I appreciate any help. Thank you!

I have a question/consultation related to statistics to all those who are good in stats! I just want to hear some thoughts or maybe I am doing this wrong. In a moderation analyses, using a regression on SPSS, I entered my predictor and moderator variable in the first block and on the second block I entered the predictorXmoderator variable. Here's my question---both models came out significant; however, the sig F change was NOT significant. Although the predictor was no longer significant in the second model but the interaction was also not a significant. Since the Sig F change from model one to model two was NOT significant, does this mean the moderation did not occur? But here's my other question, the relationship was weakened since the predictor was no longer significant in the second model after entering the interaction between the predictorXmoderator variable. Does this make sense? So was there a partial moderating effects that happened?

---

Members don't see this ad.

 

[MCParent]

You entered the interaction term after regressing it out of the parent variables, right? Or you used centered terms?

 

[grenas]

You entered the interaction term after regressing it out of the parent variables, right? Or you used centered terms?

Thanks for responding. I used centered terms.

 

[grenas]

You entered the interaction term after regressing it out of the parent variables, right? Or you used centered terms?

And yes, I entered it after regressing it out.

 

[Ollie123]

If I understand correctly, you simply have no evidence of moderation. I'd wager one of the two main effects (predictor or moderator) is highly significant and driving the overall model to significance. You see no change in the significance of the model by adding an interaction term. The interaction term is not significant. Thus, no moderation.

This community has been supportive and I've searched posts regarding moderation but did not find what I needed. Please be kind to me. I appreciate any help. Thank you!

I have a question/consultation related to statistics to all those who are good in stats! I just want to hear some thoughts or maybe I am doing this wrong. In a moderation analyses, using a regression on SPSS, I entered my predictor and moderator variable in the first block and on the second block I entered the predictorXmoderator variable. Here's my question---both models came out significant; however, the sig F change was NOT significant. Although the predictor was no longer significant in the second model but the interaction was also not a significant. Since the Sig F change from model one to model two was NOT significant, does this mean the moderation did not occur? But here's my other question, the relationship was weakened since the predictor was no longer significant in the second model after entering the interaction between the predictorXmoderator variable. Does this make sense? So was there a partial moderating effects that happened?

 

[CheetahGirl]

If I understand correctly, you simply have no evidence of moderation. I'd wager one of the two main effects (predictor or moderator) is highly significant and driving the overall model to significance. You see no change in the significance of the model by adding an interaction term. The interaction term is not significant. Thus, no moderation.

I agree that perhaps there's no moderation from your proposed model, but perhaps you could look at the significance as a function of Multicollinearity, meaning all the factors gave you significance in one pathway, but for some reason (to be stated in your discussion), the moderation model did not have significance as whole.

And remember, sometimes lack of significance is also revealing about your subject matter.

Geezzz...I also want to celebrate you being at this stage....Lucky!

 

[AcronymAllergy]

If I understand correctly, you simply have no evidence of moderation. I'd wager one of the two main effects (predictor or moderator) is highly significant and driving the overall model to significance. You see no change in the significance of the model by adding an interaction term. The interaction term is not significant. Thus, no moderation.

Ollie is much more stats-savvy than me, but this was my reading of what you'd posted as well. Sounds like there's just no moderation happening.

 

[MCParent]

If I understand correctly, you simply have no evidence of moderation. I'd wager one of the two main effects (predictor or moderator) is highly significant and driving the overall model to significance. You see no change in the significance of the model by adding an interaction term. The interaction term is not significant. Thus, no moderation.

That's my read of it as well. Sound like OP is wondering about both models being significant, but Ollie is probably right that it's just that there's one good predictor. A failed moderator wouldn't make the model nonsignificant typically.

 

[member978]

I'd agree, it sounds to me there is no evidence of moderation if the interaction term is not a significant predictor in the second step of your analyses. You might consider checking out Andrew Hayes's "Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach" for more detailed information about specific interpretations.

If you are using SPSS or SAS, after reading up on what exactly you need you might also consider looking into Hayes's macros (I think his PROCESS macro would help here) for mediation and moderation testing. You can get some pretty useful information from the macro, including data for constructing plots of the moderation, which might help illustrate the effect you are testing.

 

[grenas]

Thank you so much for all your responses! It all makes sense! I really appreciate all your help, especially Ollie! Thanks!

 

[grenas]

If I understand correctly, you simply have no evidence of moderation. I'd wager one of the two main effects (predictor or moderator) is highly significant and driving the overall model to significance. You see no change in the significance of the model by adding an interaction term. The interaction term is not significant. Thus, no moderation.

Thank you! I think I am also now in denial that my proposed moderation in the model did not happen... but that's ok!

 

[grenas]

I agree that perhaps there's no moderation from your proposed model, but perhaps you could look at the significance as a function of Multicollinearity, meaning all the factors gave you significance in one pathway, but for some reason (to be stated in your discussion), the moderation model did not have significance as whole.

And remember, sometimes lack of significance is also revealing about your subject matter.

Geezzz...I also want to celebrate you being at this stage....Lucky!

Thanks----I am hoping to defend my dissertation before I go off on internship!

 

[AcronymAllergy]

Thank you! I think I am also now in denial that my proposed moderation in the model did not happen... but that's ok!

If you found a significant effect at all, consider yourself a member of a select and fortunate subset of all dissertation defenders worldwide.

 

[cara susanna]

Yes, PROCESS is awesome for mediation/moderation analyses. And it's so easy to use, too!

Let me add to the congrats on being so far with your diss, grenas