Hi

I have a question regarding the statistics behind NBS.
In my understanding, NBS provided a solution to weakly control the multiple comparison family wise errrors in t-test between groups of edges.

I think I have a similar problem: I have a brain lesion patient cohort that have both functional and structural networks, and I'm trying to find the damaged functional edges that are significantly associated with damaged structural edges, which is tested by correlation in across the group.
Do you think the significance of the correlation can be adjusted for multiple comparison using similar approach of NBS?

Another question is regarding the packages, I found the SPC button in the NBS package, but it says it's not implemented yet, may I ask is it implemented now? If yes, where can I find the latest version of NBS?

Thank you

Yiran

Hi Yiran,
this is an interesting and novel idea. The NBS could be used to adjust for multiple comparisons, however you would need to adapt the code to suit this purpose. The current software implementation does not cater for multiple modalities (structure & function). An alternative approach would be to run the NBS separately and independently for each modality and then test for overlap in set of "damaged" functional and structural connections.

The SPC is not implemented yet. The most up to date toolbox is the one listed on NITRC. Other software implementing the NBS include CONN, MRtrix and GraphVar.


Andrew


Originally posted by YIRAN WEI: