open-discussion
open-discussion > RE: CONN report analysis-what to take
Jul 25, 2015 08:07 PM | Alfonso Nieto-Castanon - Boston University
RE: CONN report analysis-what to take
Dear Constantin,
Typically seed-to-voxel connectivity analyses are thresholded and reported in a way identical to other voxel-level functional (non-connectivity) analyses, using a combination of a "height threshold" (applied to the voxel-level statistics) and a "cluster size threshold" (applied to the different clusters of suprathreshold voxels surviving the original height threshold). A commonly used valid (appropriately corrected for multiple comparisons) combination of thresholds would be to choose a p<.001 uncorrected height threshold combined with a p<.05 FDR-corrected or FWE-corrected cluster-size threshold. When reporting these results one typically displays the suprathreshold results (either on the volume or projected to the brain surfaces) and reports the choice of thresholds used (e.g. something like "family-wise false positive levels controlled at cluster-level p<.05 FDR-corrected (height threshold p<.001 uncorrected)")
Hope this helps
Alfonso
ps. for questions about the CONN toolbox please use the CONN forum at the NITRC site ( http://www.nitrc.org/forum/forum.php?forum_id=1144 ; or go to http://www.nitrc.org/projects/conn and then select 'support' or 'forums')
Originally posted by Constantin Tuleasca:
Typically seed-to-voxel connectivity analyses are thresholded and reported in a way identical to other voxel-level functional (non-connectivity) analyses, using a combination of a "height threshold" (applied to the voxel-level statistics) and a "cluster size threshold" (applied to the different clusters of suprathreshold voxels surviving the original height threshold). A commonly used valid (appropriately corrected for multiple comparisons) combination of thresholds would be to choose a p<.001 uncorrected height threshold combined with a p<.05 FDR-corrected or FWE-corrected cluster-size threshold. When reporting these results one typically displays the suprathreshold results (either on the volume or projected to the brain surfaces) and reports the choice of thresholds used (e.g. something like "family-wise false positive levels controlled at cluster-level p<.05 FDR-corrected (height threshold p<.001 uncorrected)")
Hope this helps
Alfonso
ps. for questions about the CONN toolbox please use the CONN forum at the NITRC site ( http://www.nitrc.org/forum/forum.php?forum_id=1144 ; or go to http://www.nitrc.org/projects/conn and then select 'support' or 'forums')
Originally posted by Constantin Tuleasca:
Dear all,
I just started some seed-based analysis with CONN, it is really nice.
Could you please tell me what should be reported in an article, please:
- should take the p corrected I suppose
- than: cluster k-value, cluster p-FWE, cluster pFDR, cluster p uncorrected, peak
I found nice explanations on some pages and books but still dispatched infos on what and how to report
Thanks a lot,
Constantin
I just started some seed-based analysis with CONN, it is really nice.
Could you please tell me what should be reported in an article, please:
- should take the p corrected I suppose
- than: cluster k-value, cluster p-FWE, cluster pFDR, cluster p uncorrected, peak
I found nice explanations on some pages and books but still dispatched infos on what and how to report
Thanks a lot,
Constantin
Threaded View
| Title | Author | Date |
|---|---|---|
| Constantin Tuleasca | Jul 22, 2015 | |
| savannah ard | Jul 27, 2015 | |
| Alfonso Nieto-Castanon | Jul 25, 2015 | |
| Constantin TULEASCA | Jul 27, 2015 | |