Hello Craig:

The error degrees of freedom for the full model are calculated as follows:

dfF = 360 (# time points)
- 3*3 (3 runs, 3 baseline parameters each)
- 3*11 (3 stimuli, 11 parameters each)
- 6 (6 motion parameters)
- 3*10 (3 runs, nfirst = 10 for each run)
= 282

Note that nfirst, the first data point used in the analysis, applies to each
individual run. By default, nfirst = max of maxlags = 10. So, the first 10
data points from each run are ignored. You can get back some (or all) of these
data points by using the "-nfirst" option to override the default setting.

Also, note that there are 6 motion parameters (not 18). Since the estimated
motion parameters are entered as stim files, they are not fitted separately
for each run. To fit the motion parameters separately for each run, you
would have to enter 18 stim files (or 18 columns). That is, the 6 motion
parameters would have to be estimated (3dvolreg) separately for each run.
The motion parameter .1D files would then have to be "zero padded" in the
appropriate locations, so that each of the 18 files (or 18 columns) would be
of length 360.

Doug Ward