 |
 |
Dear Yann, I agree with Marko, in general you should go with one SPM (sub)version for one study. There have also been some updates within "final" versions of SPM8 (so after beta stage) which did affect statistics to some extent. However, it seems latest SPM8 and SPM12 versions result in identical statistics, at least for standard GLMs, probably because there were no changes at all (this is going to be different if you use older SPM8 versions). I had to rerun models with SPM8 because I wanted to conduct ROI analyses with MarsBaR, which is not compatible with SPM.mat obtained with SPM12, and well, I received identical beta estimates and T statistics. So if you do obtain identical results (T values, estimated FWHM at the bottom of the results table, ...) with both versions as well and it's just about the cluster statistics, then I would go with your current SPM8 preprocessing pipeline and results and add a footnote to your table, stating that cluster statistics for F contrasts were derived with SPM12b v... The cluster statistics are definitely a benefit. Of course a reviewer might argue that you should turn to the latest SPM12 version for your whole study, but the only relevant improvement is probably the segmentation based on six tissue probability maps, but this seems to be beta still (at least there's a statement somewhere that the TPMs might still need an update). Best, Helmut