help
help > RE: first level threshold
Oct 18, 2011 10:10 PM | Alfonso Nieto-Castanon - Boston University
RE: first level threshold
Hi Sarah,
The first-level results shown in the 'first-level analysis' tab are only thresholded based on effect size (e.g. correlation coefficient values), but this is only done for display purposes and it does not support any form of statistical inference on the single-subject results (and of course no threshold is used when passing the the first-level results to the second-level analyses). In previous versions of the toolbox (conn12) we supported fixed-effect analyses of single-subject data (those would appear in the 'second-level results' tab, selecting 'fixed-effects' instead of 'random-effects'). Those analyses used Fisher approximation of the seed-to-voxel or ROI-to-ROI BOLD signal correlations to perform inferences at the level of the individual subject results (e.g. that may be used for single-subject studies), but sadly we had to drop that functionality from the last version due to time constrains (if there is enough user interest I might get around to update that functionality back into the toolbox). In any way, if you are interested in obtaining single-subject statistics (and are willing to do some matlab coding) let me know and I will let you know the specifics of how you could do that with the current setup.
Best
Alfonso
Originally posted by Sarah Buetof:
The first-level results shown in the 'first-level analysis' tab are only thresholded based on effect size (e.g. correlation coefficient values), but this is only done for display purposes and it does not support any form of statistical inference on the single-subject results (and of course no threshold is used when passing the the first-level results to the second-level analyses). In previous versions of the toolbox (conn12) we supported fixed-effect analyses of single-subject data (those would appear in the 'second-level results' tab, selecting 'fixed-effects' instead of 'random-effects'). Those analyses used Fisher approximation of the seed-to-voxel or ROI-to-ROI BOLD signal correlations to perform inferences at the level of the individual subject results (e.g. that may be used for single-subject studies), but sadly we had to drop that functionality from the last version due to time constrains (if there is enough user interest I might get around to update that functionality back into the toolbox). In any way, if you are interested in obtaining single-subject statistics (and are willing to do some matlab coding) let me know and I will let you know the specifics of how you could do that with the current setup.
Best
Alfonso
Originally posted by Sarah Buetof:
Hey,
Sorry, that might be a stupid question, but are the first level results already somehow thresholded or FWE corrected?
Thanks for helping me out
Sorry, that might be a stupid question, but are the first level results already somehow thresholded or FWE corrected?
Thanks for helping me out
Threaded View
| Title | Author | Date |
|---|---|---|
| Sarah Buetof | Oct 18, 2011 | |
| Alfonso Nieto-Castanon | Oct 18, 2011 | |
| Sarah Buetof | Oct 19, 2011 | |
| Alfonso Nieto-Castanon | Oct 24, 2011 | |
| Kaitlin Cassady | Sep 17, 2019 | |
| Kaitlin Cassady | Oct 4, 2019 | |
| Bastien Guerin | Feb 6, 2019 | |
| Matthew Heard | Jan 22, 2019 | |
| Erin Mazerolle | Jan 17, 2014 | |
| Sarah Buetof | Oct 31, 2011 | |