We are unsure of how to determine which correction data to use. For example, should we input data reported as having small volume corrections (SVC) into SDM?

1. If the authors report both FWE and FDR corrected data, and one is significant and the other is not, which one would we choose?

2. How should we pick regions if there is a difference in results between cluster-based versus voxel-based significance?

Regarding inclusion of SVC findings:

Findings from small volume corrections should be generally discarded, unless they are also significant at the whole brain level. Let's imagine that the authors used a threshold at the whole brain level which is equivalent to t > 4.3. Then, you should include all SVC peaks with t > 4.3, whereas none with t < 4.3. This way, you are applying the same threshold to all brain regions.

Regarding selection of findings derived from different thresholds:

The more data you input to SDM, the better it recreates the brain effect-size maps. Thus, you should select the threshold with more findings – as far as it is applied to the whole brain and not to only some regions.

PS: regarding inclusion of SVC findings, note that some studies use cluster-based statistics, and thus the statistical threshold is based on cluster extent or mass rather than on peak Z value.