Hello,

I used the NBS algorithm to calculate the difference in resting state network connectivity in two different groups of subjects, which I'll call group A and group B, with group A arranged in the smaller z-dimension indices of the connectivity matrix. My contrast is set to [-1,1], and my design matrix is set up as such:

1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1
0 1

I have set a threshold of 3.3.  I get a nice significant network when I run the algorithm on this.  Based on the example from the manual, I interpreted this result to mean that there is decreased network connectivity in group B.  When I switch the contrast to [1,-1], there is no significant result, which I interpreted to mean that there is no decreased network connectivity in Group A in comparison to Group B.

However, whenever I perform calculations on the networks (using standard BCT functions) of Group A and Group B separately and compare the two groups, I'm a bit confused by the results.  My results for mean node degree, mean cluster coefficient and mean path length are all higher for group B than for group A.  I may be thinking of this wrong, but wouldn't I expect that I would have increased connectivity of Group B compared to Group A, based on these measures?  In addition, if I run group A or group B through the one-sample NBS algorithm, I get a much denser set of connections at every threshold for group B over group A.  

It seems as though I am not interpreting these results correctly -- either I don't understand what the "decreased connectivity" of group B from group A is, or it is actually an increase in connectivity in group B compared to group A.

Hi Danielle,

 

I assume that the first column of your design matrix models
Group A, and thus your connectivity matrices are arranged such that the first
16 connectivity matrices correspond with individuals from Group A. If this is
the case, then the contrast [-1, 1] will test for an increase in connectivity
in Group B (i.e. Group B > Group A).

 

The contrast [1, -1] will test for a decrease in
connectivity in Group B (i.e. Group B < Group A).

 

So it seems that you might simply be interpreting the
contrast in the opposite (wrong) direction, and thus your findings are in fact
consistent with the network measures you have computed with BCT.

 

Andrew



Originally posted by Danielle Farrar: