help
help > RE: Cluster-forming threshold for 1st l. analysis
Feb 4, 2016 02:02 PM | Alfonso Nieto-Castanon - Boston University
RE: Cluster-forming threshold for 1st l. analysis
Dear Johann,
None of the first-level volumes (separate for each subject/condition) produced by CONN are thresholded in any way (the p_corr and pFDR images simply contain the uncorrected and FDR-corrected p-values, respectively) so if you want to display/report this then, yes, you could use imcalc for example to threshold those at your desired voxel-level threshold value. For example, to display the correlation values but thresholded at p-uncorrected < .001 you could enter into imcalc both the corr_* volumes and the p_corr* volumes, and enter an expression "i1.*(i2<.001)" (that will keep in your output volume the "i1" correlation values, but only within the "i2<.001" p<.001 clusters)
Hope this helps
Alfonso
Originally posted by Johann Philipp Zöllner:
None of the first-level volumes (separate for each subject/condition) produced by CONN are thresholded in any way (the p_corr and pFDR images simply contain the uncorrected and FDR-corrected p-values, respectively) so if you want to display/report this then, yes, you could use imcalc for example to threshold those at your desired voxel-level threshold value. For example, to display the correlation values but thresholded at p-uncorrected < .001 you could enter into imcalc both the corr_* volumes and the p_corr* volumes, and enter an expression "i1.*(i2<.001)" (that will keep in your output volume the "i1" correlation values, but only within the "i2<.001" p<.001 clusters)
Hope this helps
Alfonso
Originally posted by Johann Philipp Zöllner:
Dear Conn-Pros,
when looking at the correlation coefficients in the Conn GUI (1st level), I get rather nice correlation maps with the arbitrary threshold of 0.25. But while these are nice, they are of course not reportable. When I perform the first level analysis, I get huge clusters in the p_corr and pFDR images. I *suppose* this has something to do with the cluster forming threshold that Conn uses to produce the corr images. So what I want to achieve is to adjust this threshold (I assume it is 0.001) for my 1st level 1-sample T-test(s). Alternatively, would it be correct to threshold the corr images using spm_imcalc like "i1<0.0001" (for example) to achieve the same effect?
I can provide code samples or images of my results if that is of any help!
Thanks in advance,
Johann Philipp
when looking at the correlation coefficients in the Conn GUI (1st level), I get rather nice correlation maps with the arbitrary threshold of 0.25. But while these are nice, they are of course not reportable. When I perform the first level analysis, I get huge clusters in the p_corr and pFDR images. I *suppose* this has something to do with the cluster forming threshold that Conn uses to produce the corr images. So what I want to achieve is to adjust this threshold (I assume it is 0.001) for my 1st level 1-sample T-test(s). Alternatively, would it be correct to threshold the corr images using spm_imcalc like "i1<0.0001" (for example) to achieve the same effect?
I can provide code samples or images of my results if that is of any help!
Thanks in advance,
Johann Philipp
Threaded View
| Title | Author | Date |
|---|---|---|
| Johann Philipp Zöllner | Feb 3, 2016 | |
| Alfonso Nieto-Castanon | Feb 4, 2016 | |
| Johann Philipp Zöllner | Feb 11, 2016 | |
| Alfonso Nieto-Castanon | Feb 12, 2016 | |