Dear SPM users,
I would like to do an image-based meta-analysis in SPM, that is, I want to use the full statistical images of the separate studies for the meta-analysis (see e.g. Salimi-Khorshidi et al, 2009, NeuroImage). If I understand it correctly, this can be done in SPM by creating a three-level hierarchical model, with individual subjects at the first level, group analyses at the second level, and the meta-analysis at the third level (by taking the contrast images from the second level). I have two questions about this:
- First of all, is this indeed the way to do a valid random-effects meta-analysis in SPM?
- Alternatively, to do a fixed-effects meta-analysis, would it be a valid approach to use the studies as separate groups in a single second-level analysis?
Thanks in advance for your help,
Jelmer Borst