Hi Marwa,

1) Note that the deconvolution (3dTfitter) and re-convolution will
be done one run at a time. So maybe the timing_tool.py command
could take just one run as input, or otherwise, divide the output
per run before continuing.

2) Yes, multiply the output from step 1 with the upsampled output
from 3dTfitter.

3) The TR should be that of the input, which is 0.1 at that point.

4) Yes, in this step you downsample back to the original TR grid,
so that this matches the other regressors when you give it to
3dDeconvolve (after catenating these regressors across the runs).

- rick