open-discussion
open-discussion > RE: cPPI vs gPPI
Feb 18, 2015 10:02 PM | Alex Fornito
RE: cPPI vs gPPI
Dear David,
Sorry for the delay (I did not see this post).
There are a few potential reasons for this discrepancy. Perhaps the biggest difference is the model. The model in gPPI includes PPI terms for all relevant task regressors, whereas cPPI only includes the PPI term for the contrast of interest only. To allow a more equivalent comparison, you would probably need to tweak cPPI to use a model more similar to gPPI.
Obviously, the directionality is also an issue, so you would need to see what the correlations are going in both directions.
Regards
Alex
Originally posted by David Coynel:
Sorry for the delay (I did not see this post).
There are a few potential reasons for this discrepancy. Perhaps the biggest difference is the model. The model in gPPI includes PPI terms for all relevant task regressors, whereas cPPI only includes the PPI term for the contrast of interest only. To allow a more equivalent comparison, you would probably need to tweak cPPI to use a model more similar to gPPI.
Obviously, the directionality is also an issue, so you would need to see what the correlations are going in both directions.
Regards
Alex
Originally posted by David Coynel:
Dear Dr Fornito, thanks for sharing this nice
connectivity tool. I have a general question about the
comparability of your approach with the standard PPI or the
generalized PPI (http://www.nitrc.org/projects/gppi/). I
understand that with the latter you create an asymmetric
connectivity matrix, which is not always easy to interpret in terms
of functional connectivity.
I ran both a cPPI and a gPPI in several subjects for a set of seeds, and was expecting to get connectivity values that were well correlated between the two methods. However what I observe is that within a given subject the values are rather anticorrelated, meaning that high cPPI correlation values are usually associated with lower gPPI beta estimates.
I ran both analyses using the standard scripts, so I hope I did not do anything wrong, and I would be glad to hear your opinion on that.
David Coynel
I ran both a cPPI and a gPPI in several subjects for a set of seeds, and was expecting to get connectivity values that were well correlated between the two methods. However what I observe is that within a given subject the values are rather anticorrelated, meaning that high cPPI correlation values are usually associated with lower gPPI beta estimates.
I ran both analyses using the standard scripts, so I hope I did not do anything wrong, and I would be glad to hear your opinion on that.
David Coynel
Threaded View
| Title | Author | Date |
|---|---|---|
| David Coynel | Oct 21, 2014 | |
| Alex Fornito | Feb 18, 2015 | |