Hi Giuseppe,

1. It should not matter. Motion correction is done to a base image.
If all images are shifted and rotated by the same amount in the
de-oblique step (which they should be), the only difference to the
correction would be the effect of possibly an extra resampling step.

2. This might make a small difference, but again I expect it not to
really matter. A stretching factor might end up giving slighly
different weights to different regions of the datasets, say. But
in the end, the difference should be small.

Note that in each case, the same transform (de-oblique or +tlrc)
is applied to every sub-brick of the EPI or statistical datasets.
Therefore one would expect basically the same relative transformation
to be needed for EPI to EPI-base alignment.

Also note that if the motion parameters are scaled or shifted by a
constant (and the data does not change), it will have no effect on
the statistics or the other beta weights.

- rick