 |
 |
Helmut, I think I may have fixed the issues with GLM Flex and peak_nii. If you can send me the output of GLMFlex, I can test peak_nii. If it doesn't work, it shouldn't be too hard for me to fix it. If there are any addition thoughts about what to include in peak_nii please let me know as well. Best Regards, Donald McLaren, PhD On Tue, Apr 18, 2017 at 10:02 AM, MRI More <[log in to unmask]> wrote: > Dear everyone, > > Per default second level analyses are restricted to voxels with available > data in all subjects. Has anyone adjusted SPM scripts or implemented a > toolbox that extends the analysis to voxels that have been covered in a > subset of subjects only? > > I'm aware of a corresponding option in GLM Flex, which allows to define a > threshold, e. g. 75 % of the subjects, and those voxels with incomplete > coverage are then estimated based on the available data, with some > adjustments to take into account the reduced number of dfs. However, I'm no > longer up-to-date with the newer versions (GLM Flex2, GLM Flex Fast), and > in the initial GLM Flex version I was unable to obtain cluster p values > with peak_nii. Accordingly, I would prefer a solution for SPM itself (e. g. > a toolbox). > > A while ago I also came across a paper by Vaden Jr. et al., dating back to > 2012, https://dx.doi.org/10.1016%2Fj.neuroimage.2012.01.123 , in which > (among others), the authors propose a "Available Case Analysis" strategy. > They also provide a toolbox at http://www.nitrc.org/projects/multimpute , > but as far as I could see based on the manual, it is restricted to what > they call "Multiple Imputation" - I'd prefer available case analysis. > > Best regards > > Helmut >