Hey Folks!
I was thinking about using 3dTcorrMap to look at global connectivity during 2 different task conditions, which occur during alternating ~2min blocks. As I understand it, 3dTcorrMap can give you the global connectivity for each voxel, given a timeseries.
My question is:
What's the best way to generate the 2 timeseries when the conditions are alternating?
I can imagine 4 basic approaches:
1) Chop the timeseries up into blocks, catonate the blocks, and run 3dTcorrMap on that timeseries.
2) Chop the timeseries up into blocks, run 3dTcorrMap on each block, and average the resulting maps.
3) Run 3dDeconvolve, outputting the cbucket dataset, and use 3dsynthesize to create a timeseries corresponding to just the individual condition regressors.
4) Run 3dDeconvolve, outputting the cbucket dataset, and use 3dsynthesize to create a timeseries corresponding everything but the individual condition regressors, and subtract this from the raw timeseries.
My gut feeling is that 4 is the best approach for the following reasons:
1 will introduce spurious correlations because of the catonation of the blocks.
2 will take a really long time.
3 doesn't take into consideration noise not accounted for by the initial regression model.
Any thoughts?
-nick