Dear Bethany
Nothing to worry about. The between-region connections are in units of hertz, wheres the self-connections are unitless log scaling parameters. These are transformed to units of hertz within the model, by the equation:
A_hz = -0.5 * exp(A)
Whatever value the self-connection takes (A), it will be transformed to a positive number by taking the exponential, which then scales the default self-connection of -0.5Hz. The resulting self-connection (A_hz) will always be negative.
Best
Peter
-----Original Message-----
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> On Behalf Of Bethany Sussman
Sent: 05 May 2021 01:37
To: [log in to unmask]
Subject: [SPM] resting state CSD DCM Non-Negative self connections
Dear SPM experts,
I've inverted some fully-connected (resting state, 7 nodes) one-state CSD dcms. From SPM documentation and articles, it seems that all self connections should be forced to be inhibitory connections. However, in many of my subjects there are one or more self connections that are postitive (sometimes, very positive). Is there an explanation for this or a way I can prevent it from happening?
If it makes a difference, the models are in subject space and some of the nodes are subcortical.
I've seen a few other posts about this, but I don't understand a clear answer. I am using DCM12.5 and the most up to date version of spm (7771).
Thank you in advance for any help,
Bethany
|