Dear List,
I am trying to empirically use data to determine the minimum number of
responses needed to get acceptable levels of activation. For example say
that I have two conditions (e.g. Condition 1 and 2) with 4 stimuli per
condition. Initially I modeled the data in the design matrix as follows and
then determined the number of significantly activated voxels in my favorite ROI:
Condition 1 - [3s 10s 15s 22s]
Condition 2 - [7s 12s 25s 29s]
Contrast [1 -1] or [-1 1]
I thought that this would be equivalent to modeling each stimulus as an
individual condition, then performing a combined contrast and counting the
number of significantly activated voxels in my favorite ROI.
Condition1 -3s
Condition2 -10s
Condition3 -15s
Condtion4 -22s
Condtion5 -7s
Condtion6 -12s
Condition7 -25s
Condtiion7 -29s
Contrast [.25 .25 .25 .25 -.25 -.25 -.25 -.25] or [-.25 -.25 -.25 -.25 .25
.25 .25 .25]
However when I do this, I get different results for the two different
models. I thought the two were equivalent :( If someone could explain why
they aren't equivalent could you please provide an explanation?
Second, does this mean that I need to model the first 5, 10, 15 scans etc.
that contain each of the conditions to get the minimum number of responses?
Thank you in advance.
Steve
|