Hello,
I would like to re-run my regression analyses with glt contrasts with time-series data derived from only the second and third of my three, 224-TR runs in the experiment.
I realize I can go all the way back to pre-processing and simply make a new time-series BRIK from the get-go but that seems wasteful of space and time. Better to write to file only an -fim bucket dataset for exploratory purposes.
I tried writing and using a single -censor file with 224 zeroes followed by 448 1's, but got 3Dconvolve errors when 3dDeconvolve referenced the censor file. I then tried changing concat file (specifying the beginning of runs) from
0
224
448
to
0
224
since the censor file essentially eliminated exactly one run.
THen, of course I got a glt error with bad matrices becuase my -glt 1 matrix files needed to be stripped of two zeroes each at the front to eliminate baseline terms for the missing run. I made an alternate set of matrices.
However, I still get "3Ddeconvolve Error: Improper X matrix (cannot invert X'X)
All my original event stimulus files from the full time series, and even within each run are dintinct from another, and never changed, so why collinearity?
Any idea why I am stuck? Was there a better way of going about this?