Hi All,
The reason the sequence optimization program does not return a blocked sequence is
that a blocked
sequence is optimized for detection (ie, yes or no, greater than or less than)
whereas the optimization
program attempts to minimize the expected RMS error in the estimate of the
hemodynamic response
(note that when delta functions are used as basis functions, the expected RMS
error is independent of
the actual hemodynamic response). Put another way, the stimuli are scheduled so
that the reconstructed (ie, deconvolved) hemodynamic response is as faithful to
the actual response as possible, given the inherent
noise, TR, number of samples, etc.
While blocked sequences are optimal for detection, they are terrible for trying
to
reconstruct the hemodynic response itself, and so optseq will not return block
sequences. If a blocked
deisgn is acceptible within your experimental constraints and assumptions, you
should
use it. While blocked design is optimal for detection, it does have its
drawbacks, namely
it is susceptible to the effects of adaptation, expectation, and set; stimuli
cannot be
sorted post-hoc; and the identical experiment cannot be run with EEG/MEG.
Doug
Russ Poldrack wrote:
> I believe that optseq attempts to counterbalance trial conditions to an order
> determined by the time window and the effective temporal resolution - hopefully
> Doug can provide more details on this.
>
> russ
>
> ERIC ZARAHN wrote:
>
> > Russ,
> >
> > What constraints are on the event-related design? That is, why
> > would the program not return a blocked design (which is just a specific
> > case of an event-related design) as the most efficient
> > for a given minimal spacing between events (because without any other
> > constraints I am quite sure that this is the solution)? Just curious.
> >
> > Sincerely,
> > Eric
> >
> > >
> > > on the issue of determining an optimal trial ordering for an event-related
> > > design, Doug Greve at our center has a program (part of a larger suite of
> > > fmri analysis tools) that will iteratively search for the most efficient
> > > trial ordering and timing (based on a deconvolution model). Here is the
> > > info on how to download and install his package:
> > >
>
> --
> Russell A. Poldrack, Ph. D.
> MGH-NMR Center
> Building 149, 13th St.
> Charlestown, MA 02129
>
> Phone: 617-726-4060
> FAX: 617-726-7422
> Email: [log in to unmask]
> Web Page: http://www.nmr.mgh.harvard.edu/~poldrack
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|