Within Subjects OXO and confounds

#1
Greetings! I'm helping a friend in the medical field with his research and I have an I/O Psych background - meaning lots of regression and modelling, but very little experience in the quasi-experimental field when having a control group or a group not getting a treatment is a NOT a possibility.

- He has a within subjects OXO Design (Observation of cells #1, then Treatment, then Observation #2 of cells). Dependent variable is the 2nd obervation or the change in the number of cells from #1 to #2.

- He wants to control for 4 confounding factors (2 continuous, 2 dichotomous).

He wanted to do a regression, but he doesn't have a real independent variable. He cannot assign people to yes or no treatment. What analyses can he do to see if there's a significant change from Observation #1 to Observation #2 and control or look at the effects the 4 confounding variables have on this?
 

CB

Super Moderator
#2
He wanted to do a regression, but he doesn't have a real independent variable. He cannot assign people to yes or no treatment. What analyses can he do to see if there's a significant change from Observation #1 to Observation #2 and control or look at the effects the 4 confounding variables have on this?
The fact that he can't assign participants to groups doesn't stop him from using regression. Technically an independent variable should be manipulated, but most people use the term for unmanipulated predictors as well (predictors would be a more puristically correct term!) Unmanipulated predictor variables are perfectly acceptable in regression.

That said, regression probably isn't appropriate since we're looking at a within subjects design! What might work for him is a repeated measures ANCOVA (general linear model in SPSS), in which:

-The response variable/DV is the actual outcome variable being measured at Observations 1 and 2
-Observation time (1 or 2) is a within-subjects variable
-The 2 dichotomous controls are a between-subjects variable (presuming that their categories are mutually exclusive)
-The 4 continuous controls are covariates

More info on the technique and assumptions here:
http://faculty.chass.ncsu.edu/garson/PA765/anova.htm