Dear Stephan and Rik,
In an event-related design like this, I would have thought that you SHOULD
model the 'fixation' events explicitly. If they are not modelled, then you
are asking of each voxel the question 'can a significant amount of the
variance be explained by the modelled covariate for event A' (with a
contrast 1 0 0), whereas I would rather model them explicitly and then ask
the question 'is significantly more variance explained by the modelled
covariate for event A than for the modelled covariate for event D' (with a
contrast 1 0 0 -1). I am assuming, though, that D is an actual event (such
as the appearance of a fixation point or some other temporal cue) rather
than just an arbitrarily chosen moment between the real events.
Obviously the situation is very different from an epoch design, where, (at
least with a simple box-car) the null epochs would be modelled
appropriately by a linear combination of the other epochs, so that
inclusion of an additional column for the null events is redundant.
(Incidentally, as Stephan points out, I think that the question of checking
that all of the events are sampled reasonably evenly becomes an important
one for any study in which there is a large number of types of event.)
Best wishes,
Richard.
>Stephan -
>
>> I'm designing an event-related fMRI study with 4 conditions: A,B,C,D,
>> where D is a null event (fixation). After reading Josephs and Henson
>> (1999) I still am not clear on 2 points:
>>
>> 1) If one wishes to have be sensitive to both main and differential
>> effects, is it critical to order the conditions such that each condition
>> (A,B,C,D) follows the other with equal frequency (i.e., a
>> pseudorandom, permuted design)?
>
>It is not critical; it is, in theory, slightly better for a differential
>effect
>(eg [1 -1 0] or [1 1 -2] contrast) when you are in an intermediate SOA
>(ISI)
>regime of ~6-10s (and employ the usual bandpass filter). Note condition
>D - the
>fixation trials - are not modelled. Thus a "main effect" can be viewed
>simply
>as the differential effect between each condition and fixation; the
>distinction
>here applies simply to modelled versus nonmodelled event-types.
>
>
>> Put another way, what is the disadvantage to a simple randomized design +
>> a null event for temporal jittering?
>
>No disadvantage - for short SOAs (as in your case) and null events, a
>fully randomised design is best for sensitivity to differential and main
>effects,
>and is probably what you should use.
>
>
>> 2) I have a TR= 2 seconds and a fixed ISI of 3.5 seconds. I haven't seen
>> any discussion of whether it is important to have each condition equally
>> represented at each peri-stimulus time point, but intuitively this would
>> seem to be important. For example, in this design there are 4
>> peri-stimulus positions: 0 (onset of stimulus coincides with scan), +.5
>> (stimulus presented at .5 sec after scan), -.5, and +1(same as -1).
>>
>> Shouldn't the A condition (for example) be equally represented at each
>> peri-stimulus position, to avoid bias? Apologies if this has been covered
>> previously.
>
>In theory, it is best to have each condition equally represented at each
>time point. In practice, we often assume this will be close to true with
>enough randomised stimuli.
>
>Note that in your case, I think there are 7 peristimulus sampling
>points:
>
> 0 0.5 1.0 1.5 2.0 2.5 3.0
>
>ie, you have an effective sampling rate of 2Hz (with a cycling period of
>14s: 4 ISIs or 7 TRS).
>
>Good luck!
>Rik
>
>---------------------------8-{)}-------------------------
>
>DR R HENSON EMAIL [log in to unmask]
>Wellcome Department of
> Cognitive Neurology TEL (work1) +44 171 833 7483
>12 Queen Square TEL (work2) +44 171 833 7472
>London, WC1N 3BG FAX +44 171 813 1420
>
> URL: http://www.psychol.ucl.ac.uk/rik.henson/index.html
>
>---------------------------------------------------------
from: Dr Richard Perry,
Clinical Research Fellow, Wellcome Department of Cognitive Neurology,
Darwin Building, University College London, Gower Street, London WC1E 6BT.
Tel: 0171 504 2187; e mail: [log in to unmask]
Pager: 04325 253 566.
