help > Between-subject contrast 'Any effect'
Showing 1-3 of 3 posts
May 9, 2020 05:05 PM | Mickael Tordjman
Between-subject contrast 'Any effect'
Hi Alfonso,
I have been using Conn for a few months, it is an amazing tool, thank you!
I have 2 questions regarding the "Between-subject contrast" in the Seed-to-Voxel analysis.
I have noticed that for comparisons between 2 groups ("patients" and "controls"), I can see some obvious visual differences in the Result previews when I select "Any effect (F-test)" in Between-subject contrast (image attached, subjects on the right). It seems like there is an obvious decreased connectivity in subjects compared to controls.
However, I don't find the same result when I compare patients and controls (Patients>Controls in Between-Subject contrast) with only small area of increased connectivity in patients.
What is the test performed if I select 2 groups and "Any effect" as between-subject contrast (in the result explorer box)?
Why the visual differences found when selecting "Any-effect" are not found when I select "Patients>Controls" in Between-Subject contrast? Do you think it could be due to insufficient statistical power?
Thank you very much for your reply!
Mickael Tordjman
I have been using Conn for a few months, it is an amazing tool, thank you!
I have 2 questions regarding the "Between-subject contrast" in the Seed-to-Voxel analysis.
I have noticed that for comparisons between 2 groups ("patients" and "controls"), I can see some obvious visual differences in the Result previews when I select "Any effect (F-test)" in Between-subject contrast (image attached, subjects on the right). It seems like there is an obvious decreased connectivity in subjects compared to controls.
However, I don't find the same result when I compare patients and controls (Patients>Controls in Between-Subject contrast) with only small area of increased connectivity in patients.
What is the test performed if I select 2 groups and "Any effect" as between-subject contrast (in the result explorer box)?
Why the visual differences found when selecting "Any-effect" are not found when I select "Patients>Controls" in Between-Subject contrast? Do you think it could be due to insufficient statistical power?
Thank you very much for your reply!
Mickael Tordjman
May 10, 2020 12:05 AM | Alfonso Nieto-Castanon - Boston University
RE: Between-subject contrast 'Any effect'
Hi Mickael,
The "Any effect" contrast when selecting two subject-groups simply highlights those regions showing significant connectivity with the seed ROI in any of the two groups (i.e. Patients≠0 or Controls≠0), while the "Patients>Controls" contrast will highlight those regions showing significantly different connectivity strength between the two groups (i.e. Patients≠Controls). It is perfectly fine/common to observe "differences in significance" (e.g. an effect is significant in one group but it is not significant in a second group) that are not truly "significant differences" (e.g. the between-group differences is not significant)
Hope this helps
Alfonso
Originally posted by Mickael Tordjman :
The "Any effect" contrast when selecting two subject-groups simply highlights those regions showing significant connectivity with the seed ROI in any of the two groups (i.e. Patients≠0 or Controls≠0), while the "Patients>Controls" contrast will highlight those regions showing significantly different connectivity strength between the two groups (i.e. Patients≠Controls). It is perfectly fine/common to observe "differences in significance" (e.g. an effect is significant in one group but it is not significant in a second group) that are not truly "significant differences" (e.g. the between-group differences is not significant)
Hope this helps
Alfonso
Originally posted by Mickael Tordjman :
Hi Alfonso,
I have been using Conn for a few months, it is an amazing tool, thank you!
I have 2 questions regarding the "Between-subject contrast" in the Seed-to-Voxel analysis.
I have noticed that for comparisons between 2 groups ("patients" and "controls"), I can see some obvious visual differences in the Result previews when I select "Any effect (F-test)" in Between-subject contrast (image attached, subjects on the right). It seems like there is an obvious decreased connectivity in subjects compared to controls.
However, I don't find the same result when I compare patients and controls (Patients>Controls in Between-Subject contrast) with only small area of increased connectivity in patients.
What is the test performed if I select 2 groups and "Any effect" as between-subject contrast (in the result explorer box)?
Why the visual differences found when selecting "Any-effect" are not found when I select "Patients>Controls" in Between-Subject contrast? Do you think it could be due to insufficient statistical power?
Thank you very much for your reply!
Mickael Tordjman
I have been using Conn for a few months, it is an amazing tool, thank you!
I have 2 questions regarding the "Between-subject contrast" in the Seed-to-Voxel analysis.
I have noticed that for comparisons between 2 groups ("patients" and "controls"), I can see some obvious visual differences in the Result previews when I select "Any effect (F-test)" in Between-subject contrast (image attached, subjects on the right). It seems like there is an obvious decreased connectivity in subjects compared to controls.
However, I don't find the same result when I compare patients and controls (Patients>Controls in Between-Subject contrast) with only small area of increased connectivity in patients.
What is the test performed if I select 2 groups and "Any effect" as between-subject contrast (in the result explorer box)?
Why the visual differences found when selecting "Any-effect" are not found when I select "Patients>Controls" in Between-Subject contrast? Do you think it could be due to insufficient statistical power?
Thank you very much for your reply!
Mickael Tordjman
May 10, 2020 09:05 PM | Mickael Tordjman
RE: Between-subject contrast 'Any effect'
Great, thank you very much for your reply!
Best,
Mickael
Best,
Mickael