Using GLM or UNIANOVA, (but not REGRESSION) you can directly enter covariate interactions in the model. Just click the Model button, select Custom, and build the appropriate interaction term(s).

Using syntax you would add the terms to the /DESIGN paragraph, for example:

GLM y WITH x1 x2 /DESIGN = x1 x2 x1*x2 . GLM y BY a WITH x /DESIGN = a x a*x . GLM y WITH x /DESIGN = x x*x x*x*x.

As I've hinted in the examples, one can just as easily add a covariate-by-categorical interaction (for testing homogeneity of slopes), or perform polynomial regression.

John Bauer, Ph.D. SPSS Support Statistician

-----Original Message----- From: Hector E. Maletta [mailto:hmaletta@overnet.com.ar] Sent: Tuesday, April 20, 1999 2:21 PM To: SPSSX-L@UGA.CC.UGA.EDU Subject: Re: Interaction terms in regression

David, just create the interaction variables (e.g. COMPUTE Z= X * W) and then enter them into the regression. There is no way to create the interaction terms from inside the regression procedure.

Hector Maletta Universidad del Salvador Buenos Aires, Argentina

David Huffer wrote: > > Is there a way to calculate an interaction term for a regression equation in > SPSS? With SAS you can explicitly define the terms (e.g., model > y=x1+x2+x3+(x1*x2*x3)), but for whatever reason, SPSS does not allow it > under "regression". I used GLM and entered the IV's as covariates, which > should be the same thing -- but still I wonder why I can not enter it under > the regression functions. > > DH