Dear Johann,
I have several questions of group comparison results.
For your convenience, I attached pdf file in which I wrote the questions with some figures.
I read some documents several times but still I could not get it exactly.
Best wishes,
Saske
| Attachment | Size |
|---|---|
| files.pdf | 139.17 KB |
Forums
Answers
Hi Sake,
1. “Random Networks/Groups(FDR)” will perform FDR correction with the non-parametric derived results (p-values). Only "FDR" will perform FDR correction with the parametric p-values. These results can differ if your underlying distribution is a non-parametric distribution (e.g. no gaussian distribution), or if you did not perform sufficient permutation tests with your groups (i.e., if you only do 10 permutationd the p-values are probably not a good estimate). Maybe you could read some literature about bootstrapping and its advantages.
2. p=0.0001 would be the initial link threshold (i.e., primary threshold) that determines which edges remain significant and contribute to your graph component. You will have to report this and also say why such small p-value was used. The p value of interest for the graph component would be the right column (i.e., alpha 0.01) that you will finally have to report.
3. these are group differences. The contrast is a>b and the values are negative, telling you that group b has higher connectivity between these nodes.
best,
johann
Thank you for your reply
Dear Johann,
I really thank you for your kind reply.
Now I almost understand what I asked.
Here is an related question above you answered in 3rd question.
The group differences mean the values which were calculated by subtracting correlation values (r)s of 'group a' from those of 'group b' ?
Always, I appreciate for your help.
Best regard,
Saske
group difference
I really thank you for kind reply
Always appreciate for your kind answers.
Best,
Saske