Hi,
I'm a bit new at this, and MetaAnalyst provides is a much easier intro into actually doing meta-analyses than I expected, particularly in comparison to some of the alternatives.
However, and this may relate more to the general method than to this software, MetaAnalyst does not appear to directly support the most common study designs I encounter. If I am correct, the most common meta-analysis looks at 2 group comparisons, comparing means and standard deviations after some treatment or control/placebo (equivalent to designs amenable to simple t-tests). Any pre-treatment comparisons are ignored.
The designs I would be most interested in are 2x2 partially repeated measures analyses. Two groups are selected, pre-treatment measures are taken, a treatment or control procedure is applied, and measures are taken again from each of the groups. Study results would be evaluated by ANOVA, with the comparison of interest being either the interaction effect GROUP x TIME, or the equivalent contrast examining differences in simple changes post-pre treatment per group, either case being sensitive to the net change due to treatment.
Would it be possible to integrate analysis capability of such net changes into MetaAnalyst, with direct entry of mean, SD, and N for each of the four GROUP-TIME cells per study?
Alternatively, what pre-processing procedure of such data would be best to utilize MetaAnalyst in its current form? That is, given mean, SD, and N for all four cells, how best to combine the two TIME cells to reflect the change in each GROUP? (By “best” of course I mean least likely to cause grief from journal reviewers, while simultaneously being statistically correct.)
Thanks for your patience,
Elliot