help
help > RE: Differences in SPM vs CONN 2nd level results
Dec 16, 2019 02:12 PM | Alfonso Nieto-Castanon - Boston University
RE: Differences in SPM vs CONN 2nd level results
Dear Nadja,
in CONN the p<0.01 value is from a two-sided analysis (two-tailed statistics), so the associated T threshold is 2.9208 = tinv(1-.01/2,16)
while in xjView the p<.01 value is from a one-sided analysis (one-tailed statistics), so the associated T threshold is 2.5835 = tinv(1-.01,16)
To perform the same (two-sided) analysis in SPM you would need to define a F- contrast (instead of a one-sided T- contrast), and to get the same results in xjView's gui you would need to change the "only -" option to "all". Alternatively, to perform a one-sided analysis in CONN you would need to change the "two-sided" option in CONN's gui to "one-sided (negative)"
Hope this helps
Alfonso
Originally posted by Nadja Grundinger:
in CONN the p<0.01 value is from a two-sided analysis (two-tailed statistics), so the associated T threshold is 2.9208 = tinv(1-.01/2,16)
while in xjView the p<.01 value is from a one-sided analysis (one-tailed statistics), so the associated T threshold is 2.5835 = tinv(1-.01,16)
To perform the same (two-sided) analysis in SPM you would need to define a F- contrast (instead of a one-sided T- contrast), and to get the same results in xjView's gui you would need to change the "only -" option to "all". Alternatively, to perform a one-sided analysis in CONN you would need to change the "two-sided" option in CONN's gui to "one-sided (negative)"
Hope this helps
Alfonso
Originally posted by Nadja Grundinger:
Dear All,
I have the same problem, that I get different results in CONN, SPM12 and xjView.
In the Appendix is one example of the seed-to-voxel analysis in the Default Mode Network (mPFC) at a
voxel threshold of p< 0.01 (p-uncorrected)
Cluster threshold p< 0.05 (cluster size p-FDR-corrected)
T(16)_min = 2.92 --- k_(min)= 369
In xjView:
Type: T df: 16
Threshold
-- p value = 0.01
-- intensity = 2.5835
-- cluster size = 369
Number of clusters found: 4
Why do I get a different T-intensity value in xjView as compared to CONN? If I manually Change the T-value to 2.92, which is the same value I get in CONN, my p-value changes to p=0.0050081 and then I get the same clusters as in Conn.
I'd be very grateful if anyone could help and explain this to me. Thanks!
Nadja
I have the same problem, that I get different results in CONN, SPM12 and xjView.
In the Appendix is one example of the seed-to-voxel analysis in the Default Mode Network (mPFC) at a
voxel threshold of p< 0.01 (p-uncorrected)
Cluster threshold p< 0.05 (cluster size p-FDR-corrected)
T(16)_min = 2.92 --- k_(min)= 369
In xjView:
Type: T df: 16
Threshold
-- p value = 0.01
-- intensity = 2.5835
-- cluster size = 369
Number of clusters found: 4
Why do I get a different T-intensity value in xjView as compared to CONN? If I manually Change the T-value to 2.92, which is the same value I get in CONN, my p-value changes to p=0.0050081 and then I get the same clusters as in Conn.
I'd be very grateful if anyone could help and explain this to me. Thanks!
Nadja
Threaded View
| Title | Author | Date |
|---|---|---|
| Ines Del Cerro | May 3, 2018 | |
| Nadja | Dec 16, 2019 | |
| Alfonso Nieto-Castanon | Dec 16, 2019 | |