GLM, categorical variable, interaction

[DLK]

Hello. I am trying to figure out how to do the following in GLM
univariate in SPSS:

I have a continuous DV and a categorical IV (group A or group B) and I
want to test the interaction of another continuous variable while
controlling for the effects of several variables that include a
categorical variable (race) and an ordinal variable (education).

I am confused as to how to enter the variables (fixed vs. random) and
whether all variables that I intend to control for should be made
covariates?

Should the categorical IV be entered as fixed and the rest as
covariates? Should the continuous variable involved with the
interaction be entered as random? Or should they all be made
covariates?

Or, is there another way?

Thanks for your help.

[Bruce Weaver]

In SPSS terminology, "factor" means categorical variable and "covariate"
means continuous (or at least interval or ratio-scaled) variable.

Re fixed vs. random factors, the following is a bit over-simplified,
perhaps, but should give you the gist of it.

Fixed effect: An independent variable is a fixed effect if the levels
are deliberately chosen, and are the only levels of interest. In a
replication of the study, the same levels would be used.

Random effect: An independent variable is a random effect if the levels
are randomly selected (the subjects in an experiment, for example). In
a replication, a new set of randomly selected levels would be used.

--
Bruce Weaver
bwe...@lakeheadu.ca
www.angelfire.com/wv/bwhomedir

[DLK]

> bwea...@lakeheadu.cawww.angelfire.com/wv/bwhomedir- Hide quoted text -
>
> - Show quoted text -

If the continuous variables are entered as covariates in the GLM
univariate are they then controlled for in the analysis? If not, how
does one 'control' for variables?

Thanks again.

[Bruce Weaver]

DLK wrote:

> If the continuous variables are entered as covariates in the GLM
> univariate are they then controlled for in the analysis? If not, how
> does one 'control' for variables?
>
> Thanks again.
>

Yes, each term in the model is adjusted for all the other terms in the
model.

In the case where all factors are treated as fixed (except for
subjects), the model you get via GLM-UNIVARIATE is the same as you would
get via REGRESSION. (The difference is you'd have to compute your own
indicator variables for categorical variables if using REGRESSION.
Likewise, you'd have to compute your own product terms for any
interactions.) And in a regression model, each variable is adjusted for
all other variables in the model.