Four group contrasts for moderators

[Laura K]

Dear All,

I have a dataset which compares the impact of child abuse on multiple outcomes.  There are four abuse categories, three abuse subtypes + one control group.  What I would like to do is find out whether the relationship between two continuous outcome variables, depression and parenting beliefs, differs by level of abuse.  i.e., is the relationship between depression and parenting stronger/weaker for one of the abuse subtypes (or the control group) than for the others.

My primary advisor only uses SAS, so she can't help me with SPSS.  Any help would be greatly appreciated!  Thanks!

Laura

[Maguin, Eugene]

Laura,

There are two ways to go at this problem. One is to use

Sort cases by moderator.
Split file by moderator.
Correlations depression with parenting_beliefs.

This has to be followed up the between group tests of correlations. The
formula is in Cohen and Cohen, among other sources.

The all in one way is to construct a set of three contrasts for the
moderator and then interactions with the moderator and enter all in a
regression. Like,

Recode moderator(0=0)(1=1)(else=0) into mcomp1/
   moderator(0=0)(2=1)(else=0) into mcomp2/
   moderator(0=0)(3=1)(else=0) into mcomp3.

Do repeat mod=mcomp1 mcomp2 mcomp3/
   int=ivml1 ivml2 ivml3.
+  compute int=mod*depression.
End repeat.

Regression variables=(all)/dependent=parenting_beliefs/
   enter depression mcomp1 mcomp2 mcomp3 ivml1 ivml2 ivml3.

Gene Maguin