Dear FSL Users,
We have recently discovered a FEAT GUI issue related to the “Model setup wizard” button and the “Randomise” stats option in Higher-level analysis mode. In randomise mode the “Group” column ( as viewed via “Full model setup” ) is used to define exchangeability, as opposed to variance groups which are used by FLAME. When the Model setup wizard is used to generate either the “two groups, unpaired” or “two groups, paired” designs, the resulting Group column/design.grp will represent variance groups _not_ exchangeability blocks, even if randomise has been selected to run the analysis.
For the “two groups, unpaired” option, running randomise with this design.grp will result in a single permutation and hence no significant results. For the “two groups, paired” the resulting design.grp file will imply full exchangeability, instead of only allowing exchangeability within subject, and potentially result in under-conservative statistics.
The “single group average” option is unaffected as the resulting design.grp is fine for use with both FLAME and randomise.
If you have run an analysis using a design generated with the Model setup wizard and randomise, we would _strongly_ recommend checking the design.grp that was used, and re-running with a corrected design if necessary. Note that none of this applies to manually-created designs where the design.grp will have been defined by the user.
Kind Regards
Matthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|