Dear FSL experts,
I would like to use permutation tests to analyze an fMRI dataset comprised of 3 groups and 5 conditions (separate scans). After reading the wiki and email forum, I think I have a general understanding of how to proceed, but I was hoping for confirmation on the design and contrasts. I would like to examine within-subject, between-subject, and interaction effects in my dataset. The dataset has 25+ subjects per group, but I have created a sample design (attached) using two subjects per group. I would be grateful if someone could confirm the accuracy of the design and contrasts.
I also had a few additional questions about running the analysis:
1) Because this is a mixed design, can this analysis only be executed in PALM, and not Randomise?
2) Some of my subjects have missing conditions (scans). Can PALM accommodate designs with missing repeated-measures data, and/or an unequal number of subjects per group?
3) It is unclear to me what type of exchangeability blocks should be specified for this design, and which options should be called in the command.
4) Does the one-sample mode need to be specified if I want to define a contrast for the sample average?
5) For the 4D input file for the analysis, this should be comprised of single-subject COPE files that have been registered to standard space? I know that these can be concatenated with fslmerge, but would an alternative be to instead use the filtered_func_data file from a higher-level analysis in FEAT?
Thank you for you assistance!
Michael
|