# 3 Predictor variables to predict variance in Dependent (outcome) variable that has two levels (Pre/Post scores)?

I am doing a study that is examining a trait (mindfulness) on therapy outcome. There are three predictor variables (both measure mindfulness, measured once at pre-therapy), and 2 Outcome variables I'm looking at, which are 3 different questionnaires (total scores) taken twice at pre-therapy and post-therapy. These measure PTSD, Depression and Symptom Severity.

I want to look at if mindfulness predicts somehow the outcome of therapy (in whatever way). I'm not looking for some kind of treatment effect, but rather evidence that this trait (measured by three predictor variables) influences therapy outcome.

I've been told to run a hierarchical or stepwise multiple regression, for example: Predictor variable 1 --> step 1: Outcome variable Pre-therapy; then step 2: Outcome variable Post therapy.

But I am pretty sure with a hierarchical m.r. (in SPSS) you can't input dependent variables in the step function, but rather you take one dependent variable and look stepwise at independent (predictor) variables.Is it possible to do it the way I've described?

Another idea I had would be to somehow find the difference in the pre- to post test scores of each outcome variable. And then run multiple regressions on the various predictor variables to the dependent "Difference" variables. I'm not sure how to do this though or if it is possible? My professor doesn't like the idea. He suggests the hierarchical m.r.

Does anyone have an idea of what would be the best test to use? (all variables are continuous; there is no control group - I'm simply looking at pretreatment trait mindfulness, and seeing what the effects are on pre to post treatment for PTSD; n=60) Thanks in advance!

• So P1, P2, P3 are three predictor variables measured once pre-therapy. O1 is an outcome measured pre-therapy and O2 is the same measured post therapy. Is this accurate? – Arun Jose Jun 27 '17 at 10:16
• You should not use stepwise regression (see: Algorithms for automatic model selection). – gung Jun 27 '17 at 13:30