To follow-up....the results with higher level FEAT have been the exact same running week-by-week since I posted this, so my guess was something with the original was affected with an update.
On to my follow-up massive issue with data replication: We're running the exact same lower-level feat directories through ICA-AROMA (just renamed for organization) and the ICA results are similar (but not exact) from melodic and the classifier is similar (but not exact) from AROMA ... which is yielding different data for higher level analyses. We also ran it through the higher level analysis to see if maybe this effect wouldn't be dramatic at the group level - however, the results again are similar (but not the same). All frustrating as this started as an exercise for a student who was replicating some of my published findings and kept getting slightly different results and thought he was doing something wrong.. We broke down every step of processing and narrowed it down to melodic/ICA-AROMA. Thoughts?! Is this expected?! We only found one other thread about it, but feel the need to bring this to the community's attention knowing ICAaroma is embedded into fmriprep and at this point is highly used (we love it - except now concerned). Link to other thread : https://neurostars.org/t/ica-aroma-melodic-generate-great-variabilities-on-the-same-data/22588
Please please please help :) :) :)
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|