1

I am comparing treatment outcome to two therapeutic treatments. Specifically, I am looking at how attachment moderates the relationship between therapeutic alliance and outcome. I hypothesize that the moderator of attachment will be stronger in one treatment versus the other.

The only way I can think to test this is to do two regression models. 1) alliance and outcome to Treatment one, as moderated by attachment. 2) alliance and outcome to treatment two, as moderated by attachment. Then, compare the regression coefficients to see if they are statistically different.

Is there a way to put this all in one model?

Susan
  • 11

1 Answers1

2

This can be tested in a single model with a three-way interaction term, treatment*alliance*attachment. The three-way term can be interpreted as testing whether the two-way alliance*attachment interactions differ between the two treatment levels.

Jake Westfall
  • 12,557
  • Thank you so much for your quick response! And so how does my outcome variable factor in there? I'll have gain scores for my primary outcome measure (depression). I thought I would need to make the model with my alliance scores predicting outcome scores, moderated by attachment scores. Thanks! – Susan Aug 16 '14 at 18:06
  • The outcome variable factors in as... the outcome variable. I guess I don't understand the question. – Jake Westfall Aug 16 '14 at 22:57