I was tasked to use SPSS process for my thesis and check if participant generation (Gen X, Millenial, and Gen Z) has a moderating effect on the regression between the IV and DV. How could this work? Every video I’ve seen so far seems to use just dichotomous categorical variables such as gender. Should I dummy code each generation as 1,0 like other videos suggested or is there another way to use a non-dichotomous variable?
You create two dummy variables (one group serves as reference or comparison group). The moderated regression model will contain the IV, the two dummy variables (Dummy1 and Dummy2), and two product terms (Dummy1*IV and Dummy2*IV) to reflect the potential interaction effect. For the details see Cohen, Cohen, West, & Aiken (2003), Applied multiple regression/correlation analysis for the behavioral sciences, Lawrence Erlbaum Associates, Chapter 9.
Why must you use PROCESS? You can easily estimate the model via the UNIANOVA (or equivalent GLM) command which lets you include both metric and categorical explanatory variables. Here is an example of what the basic code would look like. (I have excluded some sub-commands that just show the default settings for certain options.)
* MENU: Analyze > General Linear Model > Univariate.
* Exit via PASTE button to generate code.
UNIANOVA DV BY Generation WITH X
/PRINT=PARAMETER
/DESIGN=Generation X Generation*X.
In the code, BY is followed by one or more categorical explanatory variables and WITH is followed by one or more quantitative (or metric) explanatory variables (often described as "continuous"). In the GUI dialog, the categorical variables are called Fixed Factors and the quantitative variables are called Covariates. Because your Generation variable has 3 levels, the F-test on the Generation*X interaction will have numerator df=2. And it will test a null hypothesis stating that the slope of the relationship between X and the DV does not depend on Generation. The PRINT sub-command I included instructs SPSS to display a table of coefficients like the one you get by default from the REGRESSION command. HTH.
I agree with both answers by Christian Geiser and Bruce Weaver , but if you really need/want to use PROCESS (which is not necessary as explained), there is a button "multicategorical" since at least version 3.5 (I did not check for even older versions), where you can tell PROCESS which variables (including moderators) are multicategorical and also which coding scheme should be used (e.g. indicator, effect...).
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS