Help with a model: can the strength of moderation be used as a variable?

I am designing a study testing the role of positive emotion regulation in the relation between social anxiety disorder (SAD) symptoms and alcohol use disorder (AUD) symptoms. I am working on the data analytic plan, and realizing I am not sure how to proceed analytically.

1. The first component of my study is relatively simple: I am using ecological momentary assessment (EMA) to assess whether, on an intra-individual (within-subjects) level, alcohol consumption moderates the relation between state SAD symptoms and state nonacceptance of positive emotions. My hypothesis is that that greater state social anxiety symptoms will precede and predict greater nonacceptance of positive emotions, but that this relation will be attenuated by alcohol consumption. In other words, I predict that alcohol consumption will weaken the relationship between state social anxiety symptoms and nonacceptance of positive emotions. Again, this is on an intra-individual level.

State SAD symptoms = IV, nonacceptance of positive emotions = DV, alcohol consumption = moderator.

The model depicting aim 1 is attached:
2. The second aim of my study is where I am getting stuck. I am trying to test whether, on a between-subjects level, the strength of the moderation effect (i.e., the extent to which alcohol consumption helps those with SAD symptoms better accept their positive emotions) predicts future alcohol-related problems, measured as a DV assessed at follow-up (i.e., not using EMA for this). In other words, I plan to test whether those for whom alcohol consumption effectively diminishes the effects of state SAD symptoms on nonacceptance of positive emotions will have more alcohol-related problems at follow-up than those for whom alcohol is not effective at diminishing nonacceptance of positive emotions.

So, re: Aim 2, is there any way to assess the "strength" or "magnitude" of a moderator and use this as a between-subjects variable?
Basically, I'm trying to ask whether strength of the moderator --> alcohol use disorder symptoms at follow-up.

How would you model this? What statistical software program would you use? Am I just missing something, and there a super-obvious answer for this? :)

I am grateful for any help you can provide! Let me know if you have any questions.


TS Contributor
Well, this is interesting, the regulation of emotions by alcohol use is operationalized through the moderation effect.
The most obvious solution would be to perform n intra-individual regression analyses with predictors social anxiety,
alcohol consumption, and the anxiety*alcohol interaction, which represents the moderation effect. You could save the
n regression coefficients as a variable and use this as predictor in the second analysis.

I am not sure about the limitations of this approach, though, for example the coefficent might be biased
by certain data patterns (e.g., no or nearly no C2 consumption in an individual, or something like that).

With kind regards

Last edited: