Tools for Those Who Summarize the Evidence Base
I am designing a meta-analysis of CBT-based programs for people with low self-esteem. I have a few questions about the design and any help would be greatly appreciated. First, I have identified 19 studies that use self-esteem outcome measures, many of whom use a single-case design using pre and post-test scores. I understand that calculating effect sizes is different for independent and dependent groups, but the studies do not list the paired-measure correlations. Would it still be appropriate to calculate d using the independent measures equation? My understanding is that the leads to less precision and a wider confidence interval.
Also. 3 studies list self-esteem outcome results for subgroups (Depressed vs. Non-depressed; depressed vs anxious; one intervention and another). I understand that the number of studies (k) is a variable used to calculate a random-effects meta-analysis, so I was wondering if it would be best to combine the means and s.d. for each study rather than list the groups separately. Ideally, I would rather list the 6 different groups independently, but that will result in a misleadingly higher k.