help
help > RE: Multiple ROIs & Multiple Comparisons Correction
Aug 29, 2022 06:08 PM | Alfonso Nieto-Castanon - Boston University
RE: Multiple ROIs & Multiple Comparisons Correction
Hi Alyssa,
It depends, if you are running SBC analyses four times (one for each subnucleus) then yes, you typically will want to apply an additional correction for multiple comparisons (e.g. choosing .05/4 for the cluster-level corrected p-value threshold). On the other hand, if you select the four seeds together and run a single SBC analysis (using as a "between-sources" contrast the default [1 0 0 0;0 1 0 0; 0 0 1 0;0 0 0 1] value) then that will not need an additional correction, as it will perform a joint F- test evaluating the presence of an effect across "any" of the four seeds.
Hope this helps
Alfonso
Originally posted by yoshi:
It depends, if you are running SBC analyses four times (one for each subnucleus) then yes, you typically will want to apply an additional correction for multiple comparisons (e.g. choosing .05/4 for the cluster-level corrected p-value threshold). On the other hand, if you select the four seeds together and run a single SBC analysis (using as a "between-sources" contrast the default [1 0 0 0;0 1 0 0; 0 0 1 0;0 0 0 1] value) then that will not need an additional correction, as it will perform a joint F- test evaluating the presence of an effect across "any" of the four seeds.
Hope this helps
Alfonso
Originally posted by yoshi:
Originally posted by Alyssa Torske:
Dear Alfonso,
When uploading a ROI (as an atlas) which contains 4 parcellated subnucliei of said ROI/ seed region (e.g., left/right, medial/lateral insula) to conduct a SBC analysis, does one need to do additional multiple comparison corrections when looking at the functional connectivity changes in each subnuclei of the insula? e.i., should I be applying a Bonferroni correction and dividing the cluster threshold on second level as: FDR 0.05/4 ?
Or does CONN already take this into account somehow? I saw that for network based statistics, when there are multiple ROIs, this is done automatically, but I was not able to find any information as to whether this is also done for SBC analyses.
I look forward to hearing from you!
Kind regards,
Alyssa
When uploading a ROI (as an atlas) which contains 4 parcellated subnucliei of said ROI/ seed region (e.g., left/right, medial/lateral insula) to conduct a SBC analysis, does one need to do additional multiple comparison corrections when looking at the functional connectivity changes in each subnuclei of the insula? e.i., should I be applying a Bonferroni correction and dividing the cluster threshold on second level as: FDR 0.05/4 ?
Or does CONN already take this into account somehow? I saw that for network based statistics, when there are multiple ROIs, this is done automatically, but I was not able to find any information as to whether this is also done for SBC analyses.
I look forward to hearing from you!
Kind regards,
Alyssa
Threaded View
| Title | Author | Date |
|---|---|---|
| Alyssa Torske | Aug 9, 2022 | |
| yoshi | Aug 19, 2022 | |
| Alfonso Nieto-Castanon | Aug 29, 2022 | |
| Alyssa Torske | Sep 1, 2022 | |