Some researchers suggests multiplying the moderator variables with the moderated independent variable (if the moderator is a metric) to check differences in the regression coefficient or loading factor. Is this even applicable?
David Kenny has done significant research in this area, and has a page here that describes some theory related to moderation based on continuous and categorical variables. http://davidakenny.net/cm/moderation.htm.
Kris Preacher also has a great page on interactions. http://www.quantpsy.org/interact/interactions.htm
I suggest you to use the "PROCESS" SPSS Macro of prof. Hayes. This macro implements mediation and moderation analysis offering a wide set of models (simple or complex). You could download the macro and the guidelines for free in the following link:
David Kenny has done significant research in this area, and has a page here that describes some theory related to moderation based on continuous and categorical variables. http://davidakenny.net/cm/moderation.htm.
Kris Preacher also has a great page on interactions. http://www.quantpsy.org/interact/interactions.htm