at the moment i'm examining the effect of an intervention on health (GHQ) scores among carers of people with dissabilities, but i'm stuck with the tests I should run..
Its a 2 x 3 (quasi experimental) design with repeated measures. Between subjects Group (experimental and control) and Within Time (T1: 1 week before, T2: 1 week after and T3: 3 weeks after intervention).
My hypotheses are:
- Carers in the experimental group indicate better health over time (after the intervention).
- Carers in the control group indicate no change of health scores over time (after intervention).
- Carers in the experimental group indicate better health over time compared to carers in the control group.
However, based on descriptive statistics, the groups differ significantly at baseline in GHQ scores (and age aswell). So I have to control for those variables, right? But i'm really stuck in choosing the right statistic tests...
I read about repeated measures ANCOVA here: http://www-01.ibm.com/support/docview.wss?uid=swg21476717
So I thought about a mixed repeated measures ANCOVA (controlling for age), and if there is a significant interaction of Group X Time, the effect of time depends on group. According to the IBM site, I dont have to bother about the differences at baseline of GHC using this technique?. Then I should run simple effects in order to examine simple main effects.
But i'm totally lost if I can test each hypothese with just this ANCOVA, or only hypothese 3.
If only hypothese 3, then for 1 and 2 I have to conduct separate ANCOVA's for post scores by group while controlling for pre score?
Sorry if my question is quite stupid, but i'm a bit lost and really dont see it anymore atm...