For a research project I compared persons with and without a specific disorder on basically every published outcome I could find. The idea was to get some sort of profile of this disorder (i.e. skills that are affected or not affected). Here's an overview how the initial data looks like
Persons with and without disorder:
- Outcome A: Effect Size 1 (Study 1)
Outcome A: Effect Size 2 (Study 2) - Outcome B: Effect Size 1 (Study 1)
Outcome B: Effect Size 2 (Study 3)
Outcome B: Effect Size 3 (Study 4) - and so on...
I conducted single meta-analysis for every outcome category (i.e. Outcome A, Outcome B, ...). A multivariate meta-analysis was not possible because correlations between outcomes weren't published and outcome categories are very different. For example outcome category A is 'counting skills' and outcome category B is 'self esteem'.
My result now looks like this:
Persons with and without disorder:
- Outcome A: Effect Size (meta-analysis)
- Outcome B: Effect Size (meta-analysis)
- and so on...
I'm wondering if I can cluster or analyze these meta-analytical results further to derive some sort of profile of this disorder? The profile should explain which outcomes are (1) affected and (2) to what degree by the disorder.
I've already did some weighting of the effect sizes using a self-made formula:
weight = (Number of studies used to conduct meta-analysis) * 1/(inverse standard error of meta-analysis)
I can use the weights and maybe combine it with Cohen's scheme to categorize effect sizes (small, medium, large) to develop some sort of profile.
However I'm not really satisfied with this method and I'm looking for a statistical more advanced approach. Also, if it helps, there is a lot of a additional data within every meta-analysis which I can use (age, IQ scores, gender distribution, ...).
Please let me know if you have any good ideas! Looking forward to it!
