help > Seed-to-Voxel Results Changing
Showing 1-1 of 1 posts
Oct 30, 2019 04:10 PM | Aaron Smith
Seed-to-Voxel Results Changing
Apologies if this has already been asked, but my searches did not
yield much. I have a fairly simple dataset where a group of
subjects striated on an impulsivity measure engage in two
conditions and I ask what the connectivity changes are between
conditions as a function of impulsivity. Additionally, I've
self-defined about 16 ROIs by drawing spheres in my between-sources
contrast.
However, upon running connectivity results I am finding that my results change drastically depending on the number of seeds/sources I select out of my list of 16 in the second-level tab. My Subject effects and conditions contrasts all remain constant, but if I select a variable number of seeds/sources and use an F-test (any effects; eye(x)), the results are highly variable. Upon reading the manual and some forums, I thought that using the F-test performed contrast analyses that were independent of one another, but my variable results do not suggest this.
For instance, seeing a significant change in connectivity between a striatal ROI and a cortex cluster using an F-test of 6 different seeds in the results explorer I went into the plot effects menu to find which regions were driving the effect. The plot effects menu suggested a bilateral effect of two regions, so I exited the results explorer and re-ran the same contrast but only using the two seeds suggested by my previous analysis and the effects are gone. Does the F-test run as an omnibus test? An omnibus model would perhaps explain why these results change as the other ROIs may be soaking up error variance.
Thank you for your assistance.
However, upon running connectivity results I am finding that my results change drastically depending on the number of seeds/sources I select out of my list of 16 in the second-level tab. My Subject effects and conditions contrasts all remain constant, but if I select a variable number of seeds/sources and use an F-test (any effects; eye(x)), the results are highly variable. Upon reading the manual and some forums, I thought that using the F-test performed contrast analyses that were independent of one another, but my variable results do not suggest this.
For instance, seeing a significant change in connectivity between a striatal ROI and a cortex cluster using an F-test of 6 different seeds in the results explorer I went into the plot effects menu to find which regions were driving the effect. The plot effects menu suggested a bilateral effect of two regions, so I exited the results explorer and re-ran the same contrast but only using the two seeds suggested by my previous analysis and the effects are gone. Does the F-test run as an omnibus test? An omnibus model would perhaps explain why these results change as the other ROIs may be soaking up error variance.
Thank you for your assistance.