I'm conducting a meta-analysis of studies that used pre-post/treatment-control designs. Most of them are mixed designs (pre-post is repeated/within; treatment-control is independent/between) with a continuous dependent variable. I'm using Morris' (2008) paper to calculate effect sizes for these mixed designs. However, some of the studies are fully within designs (pre-post and treatment-control are repeated, with treatment-control counterbalanced). I have not been able to find literature on how to calculate effect sizes for a fully within pre-post/treatment-control design. Does anyone have advice or references they can share?
3
-
$\begingroup$ Is your design what in clinical trials is called a crossover study? Everyone gets both treatment and control? $\endgroup$– mdeweyCommented Apr 9, 2018 at 14:44
-
$\begingroup$ Yes, everyone receives treatment and control. $\endgroup$– JenniferCommented Apr 9, 2018 at 15:13
-
$\begingroup$ There may be material in the Cochrane handbook handbook-5-1.cochrane.org/chapter_16/16_4_cross_over_trials.htm which is relevant then. It is not really my area so cannot offer any more direct advice. $\endgroup$– mdeweyCommented Apr 9, 2018 at 16:09
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
I'm not totally clear on what it is you are after here. Could you provide details of hat you mean by a 'fully within pre-post/treatment-control design'?
However, I would point you towards this recent paper that provides a full description of how to calculate d for different experimental designs.
https://www.frontiersin.org/articles/10.3389/fpsyg.2013.00863/full
The paper also provides a flowchart in the supplementary material that helps you arrive at a decision for which formula/ standardiser to use.
-
$\begingroup$ Lakens uses eta-squared to calculate effect sizes for 2 x 2 within ANOVAs but I'm trying to use Cohen's d (most studies lack sufficient info to calculate eta-squared). The formula I'm using is from Morris (2008) and is for 2 x 2 pre-post mixed designs (1 between, 1 within factor): d = [(Mposttreatment - M pretreatment) - (Mpostcontrol - Mprecontrol)]/SDpre I'm also calculating variance of d using a formula in the Morris (2008) paper Eq 25. I was wondering if I can use the same formulas if it's a 2 x 2 pre-post within design or if there are other equations out there. Thanks! $\endgroup$– JenniferCommented Apr 9, 2018 at 14:25