Hello Marie K. I see you are in a psychology dept. People in psychology use "moderation" as a synonym for interaction. So a "moderation analysis", as you called it, it just a regression model that includes the needed interaction terms. You just need a model like this:

Grade = b0 + b1*Esteem + b2*age + b3*sex + b4*SES +

b5*Esteem*age + b6*Esteem*sex + b7*Esteem*SES

If you estimate the model via the REGRESSION command, you will have to compute the product terms yourself. But if you estimate the model via UNIANOVA (or the equivalent GLM), you can specify the product terms as I showed above on the DESIGN subcommand.

This UCLA page shows some examples of estimating linear regression models via the UNIANOVA command. Note that you have to include PRINT=PARAMETER option to see the table of regression coefficients.

• https://stats.oarc.ucla.edu/spss/seminars/introduction-to-regression-with-spss/introreg-lesson3/

PS- I know the battle has been lost, but I really dislike the use of "moderation" as a synonym for interaction. I explain why in the attached slides. ;-)

Thank you, this is very helpful!

I come from a German speaking country, and we call it "Moderatoranalyse". I haven't heard of a German equivalent for interaction analysis, but thanks for educating me. I'll keep it in mind :)

When conducting statistical analysis for a bachelor's paper with three moderation variables, you will typically employ a regression-based approach, specifically multiple regression analysis with moderation. Here's a suggested approach to analyze the data:

Multiple Regression Analysis: Begin by conducting a multiple regression analysis to examine the relationship between your dependent variable and independent variable(s). This will help establish the baseline relationship without considering moderation effects.

Identification of Moderation Variables: Determine the three variables you hypothesize as moderators of the relationship between the independent variable(s) and the dependent variable. These moderators should be theoretically relevant and have a plausible impact on the relationship.

Interaction Terms: Create interaction terms between the independent variable(s) and each of the three moderation variables. This involves multiplying the independent variable(s) with each moderation variable.

Multiple Regression Analysis with Moderation: Perform another multiple regression analysis, including the interaction terms as additional predictors alongside the independent variable(s). This analysis will allow you to examine the moderation effects.

Interpretation: Analyze the results of the multiple regression analysis with moderation. Pay attention to the coefficients of the interaction terms. If any of the interaction terms are statistically significant, it indicates the presence of moderation effects.

Further Analysis: If moderation effects are detected, conduct post-hoc analyses or simple slope analyses to interpret the nature of the moderation. These analyses can help determine the specific conditions under which the relationship between the independent variable(s) and the dependent variable is moderated by each of the three variables.

Throughout the analysis, it is important to ensure that your data meets the assumptions of multiple regression analysis, such as linearity, independence of errors, normality, and homoscedasticity. Additionally, consider controlling for any relevant covariates or confounding variables that may impact the relationship.

However, it’s worth noting that the specific analysis may vary depending on the nature of your variables, research question, and the statistical software you are using. Consulting with your advisor or a statistician that are familiar with your research field is recommended to ensure appropriate analysis techniques are applied to your specific study.

Best of Luck.

Bruce Weaver , thanks for those slides. I also try to avoid the word "moderation" and just use "interaction." I hadn't thought of one reason is that moderation could suggest a direction. You suggest "effect moderation" as a term. Do you think keeping with just interaction is okay?

Hi Daniel Wright. Sure, I'm fine with just calling it an interaction. I'm just saying that if one wanted to use some other term, I prefer effect modification over moderation because it is neutral with respect to the nature of the interaction.

Thanks Bruce Weaver !

In a nuanced world, we might use moderation as a term for a situation in which a unit increase in the moderator variable produced a decrease in the strength of the relationship. However, it is generally used to indicate an interaction effect, regardless of direction, and so is redundant.

There are other cases where this sort of terminology is useful: synergy and antagonism describe the joint effect of two treatments (often drugs) where the effect of both is either greater or less than would be expected in the basis of their effects used alone. But unless someone is going to a) promote a term in psychology for an antagonistic effect and b) enforce it, I think we should just stick with the "interaction".

The query illustrates one problem with the use of "moderation". People think that it's some sort of arcane statistic procedure, rather than a very ordinary model specification.

Ronán Michael Conroy's comment about people thinking "moderation" analysis is something special or arcane reminded me of something that happened a few years ago. A student asked me for some advice before embarking on some moderation analyses, because he had never done that sort of thing before. Or so he thought. I say that because a year or two earlier, he took a course with me in which he worked through most of the chapters in this book by Michael Mitchell:

• https://www.stata.com/bookstore/interpreting-visualizing-regression-models/

As you can see in the TOC, many of the chapters cover regression models that include various types of interactions!

I still have students in that course work through Mitchell's book. But I now mention at some point during the term (if I remember) that a "moderated multiple regression" model (MMR) is just a multiple regression model that includes at least one interaction term. ;-)