I wish to perform a moderation analysis with a categorical moderator (not dichotomous) in SPSS using PROCESS.
How should I proceed?
Should I transform my moderator in various dummy variables, and then perform one moderation analysis per dummy variable?
Hello, Ms. Alexandra,
You may check this book for guide.
Darlington, R. B., & Hayes, A. F. (2017). Regression analysis and linear models: Concepts, applications, and implementation. New York: The Guilford Press.
All the best
PROCESS will automatically transform your variables into "dummies" (chose: model 1 for moderation), and you will have one less category compared to categories of moderator variable.
It should be recoded into dichotomous variables, but if Process does it for you that is great. However, I like to use ANOVA when I have categorical responses with more than two categories because it can model categorical responses and the entire variable at once. Also, remember that just because there is interaction/moderation by one of the categories that does not mean the entire variable is a moderator. For example, if you break up Race into several "dummy variable" categories and find that African American is a moderator it is important to remember that African-American is not a variable, but just one category. Some try to get around this by testing all of the categories simultaneously and that is pretty good if all of the categories are significant moderators independently, but what if some are and some are not? This is why in my opinion the ANOVA is a better for moderation with multiple categories.
I hope this helps,
Richard
Thank you for all your inputs.
Ilija Milovanović, do you know of any published paper performing moderation analysis using Hayes's PROCESS?
Dear Professor.. Ilija Milovanović, when process is giving one less category compared to categories of moderator variable, could you please say how to interpret. Means moderator 1 (dummy coded) is W1 in process output? I have categorical moderator (5), and the process gives four interactions and moderation. How to interpret it?
If your moderator is quantitative then you should use the Johnson-Neyman interval calculations. There are reimplementations of Andrew F. Hayes's PROCESS SPSS macro for Python and R. If you like to use Python then PyProcessMacro may help:
https://github.com/QuentinAndre/pyprocessmacro
And for R there are two packages:
processR: https://github.com/cardiomoon/processR
processr: https://github.com/markhwhiteii/processr/
If you want to use ANOVA then LibreOffice Calc might be helpful:
https://help.libreoffice.org/Calc/Data_Statistics_in_Calc#Analysis_of_Variance_.28ANOVA.29
Use dummy coding for the categorical variables. if it is dichotomous just use 0 or 1. if it is k-1 dummy variable. for example
type gender male female
male 1 1 0
female 2 0 1
others 3 0 0
I recommend to use the JASP program (free access) it has an even simpler process than PROCESS to determine the moderation that can even reinforce your hypothesis according to the level of significance, that is to say beyond the frequency inference and reinforce it with the Bayesian inference with the JASP program in a simple way (it is even better than the confidence intervals), I leave the tutorial.
https://jasp-stats.org/2020/03/12/mediation-and-moderation-analysis-in-jasp/
Ilija Milovanović may i ask you how to interpret the final output? do you know articles which explain it?
Cristian Ramos-Vera Cristian_Ramos-Vera Quiero saber si l'anàlis de moderación es posible cuando la variable dependiente mide factores.
The moderation analyses according to PROCESS and other analyses are based on regression analysis with the total scores of the variables
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication