For an experimental effectiveness (or a randomized controlled trial) study in which three intervention groups and a control group undergoing a post-test, pre-test and screening test, what would be an appropriate statistical procedure for testing the hypotheses? I would appreciate answers, suggested readings, and instructions for statistical software packages.
Edit: My design looks like this:
There are three experimental interventions. A is a group psychotherapy process based on approach a. B is a group psychotherapy process based on approach c. AXB is a group psychotherapy based in blending/integrating the approaches and techniques of a and b approaches. Hypothesis is that AXB will be more effective comparative to A and B and all will be better off compared to the control condition.
There are 2 issues in your question, which are related:
One is multiple comparison: 4 arms make a lot of possible comparisons ! You may want to indicate which arms are of interest, for your a priori hypothesse and control type I risk (closed tests, tests hierarchy or ajustements)
Two is how to analyze longitudinal data with pre-test and post test measurements. I would recommend to read (recommends ANCOVA with 2-arm, which can apply in your design if you consider pairwise comparisons, provided that overall différences are significant).
Vickers AJ. The use of percentage change from baseline as an outcome in a controlled
trial is statistically inefficient: a simulation study. Biomed Central Medical Research
Methodology 2001
Hope that helps,
Cheers
Thanks for the answers so far. I feel that I need to be more specific about the design.
My design looks like the following:
There are three experimental interventions. A is a group psychotherapy process based on approach a. B is a group psychotherapy process based on approach c. AXB is a group psychotherapy based in blending/integrating the approaches and techniques of a and b approaches. Hypothesis is that AXB will be more effective comparative to A and B and all will be better off compared to the control condition.
You should use One Way Anova and Two Way Anova or their non-parametric versions to determine whether dependent variable differentiate according to the result of three experimental interventions. Because your study includes one independent variable (categorical variable or three psychotherapy process) or two independent variable (if you analyse together experimental and control group data) and one dependent variable. And you can use independent sample t test and dependent t test or their non-parametric versions to compare whether pre-test and post-test results differentiate or not.
Dear Dr. Vosylis, I have two DV's and plan to measure them three times (pretest, posttest, screening.) I will check the book out, thanks.
- Badges
- Science topic
- Similar topics
- Automotive Engineering
- Machines