help
help > RE: Single component
Nov 30, 2017 10:11 PM | Andrew Zalesky
RE: Single component
Hi Charnaya,
This suggests that the effect is widespread/diffuse. You could order the edges comprising the significant subnetworks according to effect size (t-stat value) and progressively eliminate edges with the lowest effect size.
An alternative is to eliminate nodes (and their associated connections) from the subnetwork with degree less than a given threshold.
Andrew
Originally posted by Charanya Muralidharan:
This suggests that the effect is widespread/diffuse. You could order the edges comprising the significant subnetworks according to effect size (t-stat value) and progressively eliminate edges with the lowest effect size.
An alternative is to eliminate nodes (and their associated connections) from the subnetwork with degree less than a given threshold.
Andrew
Originally posted by Charanya Muralidharan:
Hello Dr. Zalensky,
We were trying to perform some NBS t-tests (10000 permutations) and we got significant result that has 521 edges (236 nodes) in a single component at t-test threshold of 2.4 (deg of freedom = 29). When we increased the threshold to 2.41, the significance was lost. Considering we had 286 nodes to begin with and almost all the nodes show up in our component, is there a way to reduce the size of the component? Your thoughts are much appreciated.
Thank you,
Charanya
We were trying to perform some NBS t-tests (10000 permutations) and we got significant result that has 521 edges (236 nodes) in a single component at t-test threshold of 2.4 (deg of freedom = 29). When we increased the threshold to 2.41, the significance was lost. Considering we had 286 nodes to begin with and almost all the nodes show up in our component, is there a way to reduce the size of the component? Your thoughts are much appreciated.
Thank you,
Charanya
Threaded View
| Title | Author | Date |
|---|---|---|
| Charanya Muralidharan | Nov 30, 2017 | |
| Andrew Zalesky | Nov 30, 2017 | |