I am evaluating potential moderators that may interact with type of treatment (family vs. group) to predict outcome (a continuous measure of anxiety).
The 2 types of treatments are dummy coded.
So far, I tested pre to post changes (which were significant) by using the syntax: !PRE with POST; ![PRE*](a); ![POST*](a);
I then tested for interaction with tx condtion: I calculate a difference score -- DIFF = PRE-POST and regressed DIFF on Dummy for Type of Treatment.
What I want to get at is the interaction between Type of Treatment X a series of moderators, some of which are continuous (age, severity of disorder pretreatment) and others are categorical (sex, ethnicity etc.).
I need help to figure out the syntax for this. I thought of using DIFF ON Dummy and then specify a grouping variable (e.g., sex 1=female 0=male), but then I am not sure if what I get in the two groups is the interaction or if I need to constrain the path. For example: DIFF ON Dummy -- unconstrained; get chi square; DIFF ON Dummy (1) -- constrained; get chi square and then compute the difference.
Please advise. If you can suggest the syntax to use, I would greatly appreciate it. Thank you!
Thank you for your reply. I ended up creating the interaction terms in SPSS...However, I then ended up regressing the difference score. I then used GROUPING and my categorical moderators. One significant finding that I got was for Type of Treatment by Ethnicity (latino=1; non-latino=0).
The path DIFF ON TYPE OF TREATMENT was significant for non-latinos.
I am unsure how to interpret the unstandardized coefficient for
NON-LATINO GROUP DIFFRPPO ON COND 4.196 The intercept for DIFFRPPO 2.666 COND is dummy coded FAMILY (1) vs GROUP (0).
I'm thinking that 4.196 is the difference from pre to post for the group scored 0 (i.e., GROUP).
4.196+2.666=6.862 will be the difference from pre to post for the group scored 1 (i.e., FAMILY).
Could you please advise if this is the case or the correct way to interpret this finding? After I figure it out with one moderator, I will follow the model with the other moderators I am trying out. Thank you very much.