help > punc in the review and results explorer?
Showing 1-2 of 2 posts
May 15, 2017 11:05 AM | Dayana Hayek - Charite Universität Medizin Berlin
punc in the review and results explorer?
Dear all,
I am currently using the second-level in CONN v17.a, whenever I choose a covariate and check the results:
In the preview on the right Panel I see positive correlations with p unc<0.001, however, whenever I go to the results Explorer, I see no results.
Is there any difference between these two Tools? Why would I see nice Cluster (with p values less than 0.001) in the preview and it disappears when I open the results Explorer?
Many thanks and Kind regards,
Dayana Hayek
I am currently using the second-level in CONN v17.a, whenever I choose a covariate and check the results:
In the preview on the right Panel I see positive correlations with p unc<0.001, however, whenever I go to the results Explorer, I see no results.
Is there any difference between these two Tools? Why would I see nice Cluster (with p values less than 0.001) in the preview and it disappears when I open the results Explorer?
Many thanks and Kind regards,
Dayana Hayek
May 24, 2017 03:05 AM | Alfonso Nieto-Castanon - Boston University
RE: punc in the review and results explorer?
Dear Dayana,
There are two differences in the method and nature of the thresholding operation shown in the 'preview' display and the 'results explorer' window. The first is that the 'preview' results are only thresholded using a voxel-level height threshold (e.g. p<.001) without any form of cluster-level thresholding or multiple-comparison correction like the one implemented in the 'results explorer' window (e.g. using cluster-level p-FDR<.05). The second is that, by default, the 'preview' window is showing you two-sided results while the 'results explorer' window is showing you the one-sided results. If you want to see in the 'results explorer' window exactly the same voxels/results as you were seeing in the 'preview' window simply use a liberal-enough cluster threshold (e.g. cluster-size > 0) and change the 'one-sided (positive)' setting to 'two-sided'.
Hope this helps
Alfonso
Originally posted by Dayana Hayek:
There are two differences in the method and nature of the thresholding operation shown in the 'preview' display and the 'results explorer' window. The first is that the 'preview' results are only thresholded using a voxel-level height threshold (e.g. p<.001) without any form of cluster-level thresholding or multiple-comparison correction like the one implemented in the 'results explorer' window (e.g. using cluster-level p-FDR<.05). The second is that, by default, the 'preview' window is showing you two-sided results while the 'results explorer' window is showing you the one-sided results. If you want to see in the 'results explorer' window exactly the same voxels/results as you were seeing in the 'preview' window simply use a liberal-enough cluster threshold (e.g. cluster-size > 0) and change the 'one-sided (positive)' setting to 'two-sided'.
Hope this helps
Alfonso
Originally posted by Dayana Hayek:
Dear all,
I am currently using the second-level in CONN v17.a, whenever I choose a covariate and check the results:
In the preview on the right Panel I see positive correlations with p unc<0.001, however, whenever I go to the results Explorer, I see no results.
Is there any difference between these two Tools? Why would I see nice Cluster (with p values less than 0.001) in the preview and it disappears when I open the results Explorer?
Many thanks and Kind regards,
Dayana Hayek
I am currently using the second-level in CONN v17.a, whenever I choose a covariate and check the results:
In the preview on the right Panel I see positive correlations with p unc<0.001, however, whenever I go to the results Explorer, I see no results.
Is there any difference between these two Tools? Why would I see nice Cluster (with p values less than 0.001) in the preview and it disappears when I open the results Explorer?
Many thanks and Kind regards,
Dayana Hayek