sdm-help-list > Huge clusters unaffected by threshold
Showing 1-4 of 4 posts
Apr 14, 2015 04:04 PM | Nobody
Huge clusters unaffected by threshold
Hello,
First of all, thank you so much for the great software you built and for your generosity by sharing it and providing tutorials and help.
I have run a meta-analysis on functional data related to pain on 29 studies.
After thresholding "MyMean_z" even after very conservative thresholds, I obtain very large blops/clusters of activation.
I ran the same analysis with GingerALE, I don't have so huge blops even at FDR p<.05.
Based on the location of the peaks, it looks like GingerALE maps more closely onto the location of the peaks.
See the picture attached.
Do you have any suggestion to improve the results?
Thank you.
First of all, thank you so much for the great software you built and for your generosity by sharing it and providing tutorials and help.
I have run a meta-analysis on functional data related to pain on 29 studies.
After thresholding "MyMean_z" even after very conservative thresholds, I obtain very large blops/clusters of activation.
I ran the same analysis with GingerALE, I don't have so huge blops even at FDR p<.05.
Based on the location of the peaks, it looks like GingerALE maps more closely onto the location of the peaks.
See the picture attached.
Do you have any suggestion to improve the results?
Thank you.
Apr 14, 2015 04:04 PM | Nobody
RE: Huge clusters unaffected by threshold
SMALL UPDATE:
I can of course change the threshold within MRIcron, and then looks much more accurate than in GingerALE. Cf picture attached.
However, I have arbitrarily chosen that threshold in MRIcron.
I could not get to that threshold within SDM.
Using the most conservative threshold led to the same big blobs (i.e., MyMean_z_p_0.00001_1.000_50.nii)
I can of course change the threshold within MRIcron, and then looks much more accurate than in GingerALE. Cf picture attached.
However, I have arbitrarily chosen that threshold in MRIcron.
I could not get to that threshold within SDM.
Using the most conservative threshold led to the same big blobs (i.e., MyMean_z_p_0.00001_1.000_50.nii)
Apr 21, 2015 08:04 AM | Joaquim Radua
RE: Huge clusters unaffected by threshold
Dear,
when I obtain the same blobs using very conservative thresholds, I feel safer that these are just the results. Of course, one should always consider that very significant blobs may be artificially enlarged due to the smoothing applied in the original studies and in the meta-analyses, but on the other hand this smoothing is somehow necessary to account for the spatial imprecision. However, if you feel uncomfortable with it, you can also show a map using a narrower kernel (i.e. after pre-processing the studies using a lower FWHM).
Hope this helps,
Joaquim
when I obtain the same blobs using very conservative thresholds, I feel safer that these are just the results. Of course, one should always consider that very significant blobs may be artificially enlarged due to the smoothing applied in the original studies and in the meta-analyses, but on the other hand this smoothing is somehow necessary to account for the spatial imprecision. However, if you feel uncomfortable with it, you can also show a map using a narrower kernel (i.e. after pre-processing the studies using a lower FWHM).
Hope this helps,
Joaquim