Hi Sara,

since your moderator is categorical/binary, indeed doing a multigroup model with all variables would be feasible (you don't need product terms).

What I wonder is why you have only one predictor variable. Is this an experimental treatment? If not I would invest time into thinking about potential confounders of the X-Y effect and include them.

If your have several indicators for each variables and you could defend a common factor model, going to multigroup SEM would allow to test for measurement invariance of the measurement model.

Best,

Holger

With 4 DVs I would start with a multivariate (more than 1 y not more than 1 x) regression model. Andrew Gelman has done important work in this area so I would look at this link for assistance.:

https://www.bing.com/search?q=Gelmam+Multivariate+hierarchical+regression&form=PRUSEN&pc=BDT1&mkt=en-us&httpsmsn=1&refig=41c7c2d207044f20b02b89e3b12717c7&sp=-1&pq=gelmam+multivariate+hierarchical+regression&sc=1-43&qs=n&sk=&cvid=41c7c2d207044f20b02b89e3b12717c7

Best wishes, David Booth

There is a PROCESS macro by Hayes. It works on SPSS and has advanced options in mediation analysis such as conducting bootstapping and providing a script for visualizing the output. In terms of moderation analysis, I believe it is one of the best options. the interface is very familiar to other SPSS dialogs, you can easily add confounders-covariates etc. into the model. You need to learn how to use the macro and results from documentation provided by the author. You can have a look at: https://processmacro.org

Hayes asserts that if the model do not include true latent variables(i.e. if you use summed/average scores); usage of this procedure is as acceptable as SEM.

Although you can analyze moderator variable using spss but its better to do it by structural equation modeling using a SEM softwares ( Pls - Mplus - Amos ) or open source softwares like R, jasp because of host of benefits. by using regression you need to compute your questions in order to form a variable and this will ruin your data. in the other hand, when you use SEM it will impute the data instead . of course this is not the only advantage of SEM over typical regression but an important one.

for more information check out these authors books: christian ringle, Joseph F. Hair.

hope this helps.