Jack,
> Could you tell me why SPM doesn't use temporal filter for drift and shift
> in MRI time series (2 lines in matlab and faster computation), but
> implement a matrix of undesirable data ?
With regards to "drift", including low-frequency sinusoids in the design
matrix effectively hi-pass filters the data and also simultaneously
accounts for the reduction of the dimensionality of the data vector by
doing so. Simply hi-pass filtering the data (by, for example, using a
matlab routine) would not effect the
latter, which is important for statistical inference.
I am not sure what you specifically mean by "shift"; could you
elaborate?
Sincerely,
Eric
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|