> Additionally, we've found that 6 - 9 s between repeats of the same
> condition, on average, are best for estimating a condition vs. baseline
> (mean), with random jitter.
This is completely in line with my experience. Usually, I use around 3.5-4
seconds ISI between different conditions and around 8-12 seconds ISI between
the same condition (with random jitter of course). And I've got always very
reliable and robust results in the main contrasts as well as the difference
contrasts.
Karsten
----------------------------------
Karsten Specht
fMRI Section
Department of Neuroradiology
Medical Center Bonn
Spessartstrasse 9
53119 Bonn
Germany
Phone: ++49-(0)228/90 81-178
Fax: ++49-(0)228/90 81-190
E-Mail: [log in to unmask]
WWW: http://www.mcbonn.de/Praxis/praxis15/fmri1.htm
> -----Ursprungliche Nachricht-----
> Von: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]Im
> Auftrag von Tor Dessart Wager
> Gesendet: Donnerstag, 5. September 2002 19:02
> An: [log in to unmask]
> Betreff: Re: Fw: which ISI should I use?
>
>
> > > e.g., Dale, 1999, Human Brain Mapping, and Liu
> > > et al., Neuroimage, 2001. The Dale paper shows that if you jitter the
> > > ISI then efficiency goes up as you decrease the ISI. I doubt
> that it's
> > > ever going to be the case that increasing the ISI (keeping total scan
> > > time constant) gives you greater power to detect differences between
> > > conditions (though it will increase your ability to estimate the
> > > response against baseline).
> > >
> >
> > a question - I don't have the paper by Dale et al. available -
> but is it not
> > the case that they make their efficiency-analysis under the (strong)
> > assumption that the brain-bold system behaves as a linear time invariant
> > response system? and if I understand Andre's question
> correctly, he implies
> > that non-linear effects (e.g. ceiling effects) may be
> problematic at (too)
> > short ISI when an approximate steady-state situation is reached
> - may there
> > not be some kind of trade-off here?
>
> I think this is so. If you're using 3 s as the shortest possible ISI, you
> probably don't need to worry about it. In our simulations, we've seen
> problems arise from BOLD nonlinearity when the ISI is less than 2 s, and
> some papers (by Aguirre, Friston, Vasquez & Noll, Birn[?]) seem to support
> this
> general guideline.
>
> It's true that under the assumptions of the linear model, the shorter the
> ISI, the better the power to detect differences (but not main effects!),
> but one thing that isn't considered by these simulations that's related to
> nonlinearity is scaling issues, which appear in other discussions on the
> list. Even if you have more power with a very short ISI, the relative
> rise and fall of your predictors (with a random ER design) will depend on
> the sampling resolution (i.e. default TR/16 in SPM), and they may be
> wrong. So that's an additional reason that using TR < 2 s may be a bad
> idea unless you're careful about these issues.
>
> Additionally, we've found that 6 - 9 s between repeats of the same
> condition, on average, are best for estimating a condition vs. baseline
> (mean), with random jitter. I've heard a number of people say that an
> exponentially decreasing distribution (most at 2 s, fewer at longer ISIs)
> is optimal, but I'm not sure how that was concluded or why it should be
> true. Can anyone else comment on that?
>
> Tor
>
>
>
> > > > does this rule apply to every situation? What if both conditions are
> > > > expected to activate the identical cortical area, but with a rather
> > > > slight different strength? In this case, there might be the
> danger that
> > > > the BOLD response will level off at some intermediate
> signal strength
> > > > (due to the sluggishness) and that a contrast between these
> conditions
> > > > will result in no difference. Might it be, that for such a
> hypothesis
> > > > longer ISIs are required?
> > > > Andre Szameitat
> >
> > <<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>
> > karl magnus petersson
> > Neurocognition of Language Research Group
> > Max Planck Institute for Psycholinguistics
> > Postbus 310, NL-6500 AH Nijmegen, The Netherlands
> > http://www.mpi.nl/world/index.html
> >
> > F.C. Donders Centre for Cognitive Neuroimaging
> > University of Nijmegen
> > P.O. Box 9101, NL-6500 HB Nijmegen, The Netherlands
> > http://www.kun.nl/fcdonders/website/index.php
> >
> > Cognitive Neurophysiology Research Group R2-01
> > Department of Clinical Neuroscience, Karolinska Institutet
> > Karolinska Hospital, S-171 76 Stockholm, Sweden
> >
> > Phone:+31-24-3610984/+46-8-51772039;
> > Fax: +31-24-3610652/+46-8-344146;
> > Email: [log in to unmask]
> > <<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>
> >
>
|