Dear CONN experts,
I came across a paper (Megumi et al., 2015, Frontiers in Human Neuroscience) in which the authors computed the connectivity between pairs of networks (3 networks (A B C), 6 possible pairs (AA, AB, AC, BB, BC, CC)) and I'd like to know if this approach is valid:
1. There are 3 groups of subjects and the connectivity matrices are computed 3 times (pre-intervention (pre) and post-intervention (post))
2. Regions for each network are defined based on the literature.
3. The increase of connectivity in an ROI is computed as: [i]Increase[/i] = post minus pre
4. The [i]mean [/i]connectivity and [i]SD [/i]across groups is computed for post minus pre
5. Whenever [i]Increase [/i]is higher than [i]mean + SD[/i], [i]Increase[/i] is considered significant for a specific connection
6. For each network pair (6 pairs): The number of significant connections (see 5) in a network pair is computed, as well as the CI of this number, by using the bootstrapping approach
7. Correction by multiple comparisons
8. They finally determine which network pairs exhibited a larger number of significant connections than the corrected (100 - alpha)% CI.
Does anybody can say something about its validity? If it is valid, does CONN perform something similar? I guess the authors do it because comparing average connections of network pairs would not lead to significant differences since only a few connections would be different.
Thank you in advance!
Gustavo Pamplona
I came across a paper (Megumi et al., 2015, Frontiers in Human Neuroscience) in which the authors computed the connectivity between pairs of networks (3 networks (A B C), 6 possible pairs (AA, AB, AC, BB, BC, CC)) and I'd like to know if this approach is valid:
1. There are 3 groups of subjects and the connectivity matrices are computed 3 times (pre-intervention (pre) and post-intervention (post))
2. Regions for each network are defined based on the literature.
3. The increase of connectivity in an ROI is computed as: [i]Increase[/i] = post minus pre
4. The [i]mean [/i]connectivity and [i]SD [/i]across groups is computed for post minus pre
5. Whenever [i]Increase [/i]is higher than [i]mean + SD[/i], [i]Increase[/i] is considered significant for a specific connection
6. For each network pair (6 pairs): The number of significant connections (see 5) in a network pair is computed, as well as the CI of this number, by using the bootstrapping approach
7. Correction by multiple comparisons
8. They finally determine which network pairs exhibited a larger number of significant connections than the corrected (100 - alpha)% CI.
Does anybody can say something about its validity? If it is valid, does CONN perform something similar? I guess the authors do it because comparing average connections of network pairs would not lead to significant differences since only a few connections would be different.
Thank you in advance!
Gustavo Pamplona