help > Can I use FDR and FWE correction together
Showing 1-2 of 2 posts
Nov 6, 2017 05:11 AM | Hua Yu
Can I use FDR and FWE correction together
Dear NBS experts,
Sorry to disturb you so much. I have another stupid question. I used the NBS FWE correction and found significant group difference between HCs and patients by using the contrast [1, -1]. However, there's only FDR significant results when I use the contrast [-1, 1]. So can I use different correction methods in the same group comparison analysis for different contrasts?
Many thanks,
Hua
Sorry to disturb you so much. I have another stupid question. I used the NBS FWE correction and found significant group difference between HCs and patients by using the contrast [1, -1]. However, there's only FDR significant results when I use the contrast [-1, 1]. So can I use different correction methods in the same group comparison analysis for different contrasts?
Many thanks,
Hua
Nov 6, 2017 05:11 AM | Andrew Zalesky
RE: Can I use FDR and FWE correction together
I suggest using the same method for all comparisons. It does not
make sense to cherry pick methods based on desired significance.
When using FDR, it is VERY IMPORTANT to use 50000-100000 permutations or more.
This can take a long time to run. For NBS, 5000 permutations is ok.
Andrew
Originally posted by Hua Yu:
When using FDR, it is VERY IMPORTANT to use 50000-100000 permutations or more.
This can take a long time to run. For NBS, 5000 permutations is ok.
Andrew
Originally posted by Hua Yu:
Dear NBS experts,
Sorry to disturb you so much. I have another stupid question. I used the NBS FWE correction and found significant group difference between HCs and patients by using the contrast [1, -1]. However, there's only FDR significant results when I use the contrast [-1, 1]. So can I use different correction methods in the same group comparison analysis for different contrasts?
Many thanks,
Hua
Sorry to disturb you so much. I have another stupid question. I used the NBS FWE correction and found significant group difference between HCs and patients by using the contrast [1, -1]. However, there's only FDR significant results when I use the contrast [-1, 1]. So can I use different correction methods in the same group comparison analysis for different contrasts?
Many thanks,
Hua