help > duplicate connections exporting table
Showing 1-5 of 5 posts
May 20, 2019 09:05 PM | Jen Labus - UCLA
duplicate connections exporting table
Hi Alfonso,
We are encountering duplicates in the results window (headings: Analysis Unit, Statistic, P-Unc, pFDR, PFwe).
We sort by connections and see the same connectivity but with different order for example Amygdala- OFC and OFC to Amygdala.
The duplicates remain in the exported table. Is there a way to limit the results to show a connection one time?
Thanks,
Jen
We are encountering duplicates in the results window (headings: Analysis Unit, Statistic, P-Unc, pFDR, PFwe).
We sort by connections and see the same connectivity but with different order for example Amygdala- OFC and OFC to Amygdala.
The duplicates remain in the exported table. Is there a way to limit the results to show a connection one time?
Thanks,
Jen
May 29, 2019 03:05 AM | Jen Labus - UCLA
RE: duplicate connections exporting table
HI Alfonso and group?
Has this been previously addressed?
Has any one else run int this issue?
Has this been previously addressed?
Has any one else run int this issue?
Jun 12, 2019 05:06 AM | Jen Labus - UCLA
RE: duplicate connections exporting table
Hi Alfonso,
I can live with duplicates. It makes sense that if I select all rois in the graph explorer I will get both region A to region B and Region B to Region A in the output table . However, not all the ROIs are duplicated (see 381-89) and not all of the ROIs have the same calculated FDR values (see 188 to 146 and the same connection 146 to 188).
188 146 -3.99 0.0001 0.0419
146 188 -3.99 0.0001 0.0399
271 68 -3.96 0.0001 0.0476
68 271 -3.96 0.0001 0.0476
381 89 -3.94 0.0001 0.0064
I am using conn 17a. I have enclosed two new screen shots. Why would 381 to 89 be significant but not 89 to 381?
all the very best,
Jen
I can live with duplicates. It makes sense that if I select all rois in the graph explorer I will get both region A to region B and Region B to Region A in the output table . However, not all the ROIs are duplicated (see 381-89) and not all of the ROIs have the same calculated FDR values (see 188 to 146 and the same connection 146 to 188).
188 146 -3.99 0.0001 0.0419
146 188 -3.99 0.0001 0.0399
271 68 -3.96 0.0001 0.0476
68 271 -3.96 0.0001 0.0476
381 89 -3.94 0.0001 0.0064
I am using conn 17a. I have enclosed two new screen shots. Why would 381 to 89 be significant but not 89 to 381?
all the very best,
Jen
Jun 12, 2019 05:06 AM | Jen Labus - UCLA
RE: duplicate connections exporting table
second screen shot same contrast.
Jun 12, 2019 08:06 AM | Pravesh Parekh - National Institute of Mental Health and Neurosciences
RE: duplicate connections exporting table
Hi Jen,
The way Conn applies FDR correction is seed based. For a given seed (say 381), all the target connections' p values will be considered when calculating the FDR cut-off (q value). Vice-versa when looking at 89, the cut-off will be based on the p values obtained from all the p values when considering 89 as the seed. Therefore, it is quite possible that 381-89 is significant but not 89-381 as the correction is dependent on the actual values when considering each of these as a seed.
Here is an older forum post about this: https://www.nitrc.org/forum/message.php?...
and here is a link to a detailed reply from Dr. Alfonso: https://www.nitrc.org/forum/message.php?...
Hope these help
Best
Pravesh
The way Conn applies FDR correction is seed based. For a given seed (say 381), all the target connections' p values will be considered when calculating the FDR cut-off (q value). Vice-versa when looking at 89, the cut-off will be based on the p values obtained from all the p values when considering 89 as the seed. Therefore, it is quite possible that 381-89 is significant but not 89-381 as the correction is dependent on the actual values when considering each of these as a seed.
Here is an older forum post about this: https://www.nitrc.org/forum/message.php?...
and here is a link to a detailed reply from Dr. Alfonso: https://www.nitrc.org/forum/message.php?...
Hope these help
Best
Pravesh