I've compared two independent groups using an unpaired t-test. I've used the 246 node brainnetome atlas. One group has less connectivity involving 82 nodes. I've then extracted fully weighted graph theory measures (e.g., path length) using the 82 nodes from the initial NBS result. The same group with less NBS  connectivity has significantly higher path lengths - i.e.,  less efficient information transfer across the 82 node network. 

My question: is what I've done circular or are the initial NBS and follow up graph theory comparisons independent from one another? 

Many thanks, Liam.

Hi Liam, 

I am not sure if I completely understand your analyses, but it seems that some degree of "cherry picking" is evident here. 

it seems that the subnetwork showing a between group difference is isolated and then tested for a difference in path length. It could be argued that path length will likely differ in this subnetwork because we have already established that there is a between-group difference. 

That said, graph measures can potentially provide some insight into the characteristics of the between-group difference in connectivity discovered with NBS 

best,
Andrew

Originally posted by liam nestor:

Many thanks Andrew. Just to clarify - I've found 82 nodes in the initial NBS where one group had significantly less connectivity.  I then used those nodes to extract graph theory measures from in GraphVar. Path length is longer across these nodes in the same group and the cluster coefficient is smaller. I found, however, that some other graph theory measures across these 82 nodes (e.g., small world propensity) were not different between the two groups.

I'm asking if there's evidence of circularity because a reviewer is saying that what I've done is circular and doesn't add anything new to the initial NBS result. All the best, Liam.

Hi Liam, 

yes - I understand how this could be construed as potentially circular. This is because the graph analyses are performed on the subnetwork for which we have already declared to show a between-group difference. 

On the other hand, the graph measure can give us insight into the between-group difference. 

So I think it can be argued in both ways - not really clear cut in my opinion. 

Andrew
Originally posted by liam nestor:

Thanks Andrew. I think I'll just have to make a case about how graph theory measures can provide additional insight into what might be driving the initial NBS group differences. All the best, Liam.

Originally posted by Andrew Zalesky:

Thank you Andrew sir for answering, you saved my day. I am glad I found this post. I was actually searching for this https://fitmymoney.com/500-dollar-loan/ website because I need a loan and for that I am looking for a site online through which I can get a loan easily. And while searching for it online, I found your post.