[R] Cohen's d on Adjusted Means (UNCLASSIFIED)

[Bliese, Paul D LTC USAMH]

Classification:  UNCLASSIFIED 
Caveats: NONE

I have a large group randomized trial (pre-post design) where the
randomization was marginally successful.  Given the pre-existing
differences among groups, it makes sense to report adjusted means (aka
least squares means though I estimated them via predictions from lme
rather than lm).

I would like to report effect sizes (Cohen's d) calculated as:

(Mean1-Mean2)/sqrt(((N1*VAR1)+(N2*VAR2))/(N1+N2))

Can anyone provide advice on what would be a reasonable value to use for
the variance estimate in the equation?

Any other thoughts on how to calculate effect sizes on adjusted mean
values?

Thanks,

Paul
 

LTC Paul D. Bliese
Commander, US Army Medical Research Unit - Europe
Heidelberg, Germany
DSN:  314-371-2626
COMM:  +49-6221-172626 
Classification:  UNCLASSIFIED 
Caveats: NONE