help > Output of test_stat
Showing 1-3 of 3 posts
Display:
Results per page:
Nov 28, 2018  08:11 AM | joshkr
Output of test_stat
Dear NBS team, 


Thank you for your excellent program. 


I am trying to understand the output of test_stat, which I have obtained with 'global nbs' after outputting my result (derived using FDR with a T test).


It is not clear what the numbers in the matrix represent, are they the T values of each edge? Without the significance threshold I am finding it difficult to interpret.

Assuming that they are T values, I have tried to determine the threshold used by NBS for my data, however those with the highest values do not 100% approximate to those edges that nbsview delivered. 


As a side point, despite performing 100000 permutations, I am still receiving slightly varied results, how many permutations would you suggest?


Any help would be much appreciated. 


Kind regards, 

Josh
Nov 28, 2018  05:11 PM | Andrew Zalesky
RE: Output of test_stat
Hi Josh,

test_stat stores the test statistic for the set of all edges. If you select t-test in the GUI, then the test statistic values represent t statistics.

Significance is based on permutation testing, and thus the test statistic in itself is not enough to determine significance.

For NBS, 5,000 permutations should be enough.

For FDR, a substantially greater number of permutations are required. Perhaps 100,000 or 500,000 to be safe with FDR.

Andrew



Originally posted by joshkr:
Dear NBS team, 


Thank you for your excellent program. 


I am trying to understand the output of test_stat, which I have obtained with 'global nbs' after outputting my result (derived using FDR with a T test).


It is not clear what the numbers in the matrix represent, are they the T values of each edge? Without the significance threshold I am finding it difficult to interpret.

Assuming that they are T values, I have tried to determine the threshold used by NBS for my data, however those with the highest values do not 100% approximate to those edges that nbsview delivered. 


As a side point, despite performing 100000 permutations, I am still receiving slightly varied results, how many permutations would you suggest?


Any help would be much appreciated. 


Kind regards, 

Josh
Nov 29, 2018  01:11 AM | joshkr
RE: Output of test_stat
Dear Andrew, 



Thank you very much for the prompt and clear reply. 


Kind regards, 


Josh