Dear all,
Quite possibly a stupid question, but ..
I try to identify predictors of Reliable Change using the CBCL.
In the bivariate analysis, (t-tests), there were significant
differences between those who improved reliably and those who did not
on their CBCL subscale scores at intake (e.g anxious/depressed,
somatic symptoms etc) and also on Internalising and Total scale
scores at intake.
Would it make any sense to use all those as predictors in Logistic
Regression, or would it be a problem given the fact that total
scores>internalising scores>anxious/depressed, somatic symptoms
subscales in terms of hierarchical organisation of the variables
(i.e. in order to compute internalising problems, anxious/depressed,
and somatic symptoms subscales are used and in order to compute
total problems internalising and externalising problems are used).
Am I right in guessing that multicollinearity would be an issue and
in that case one should use only the lower order variables?
Many thanks
Ioanna
|