In my humble opinion, they should NOT be modelled,
assuming that they are TRUE null events (as defined
in Josephs & Henson, 1999), ie NOT detectable by
subjects. They therefore have no definition in time.
They are just an easy practical means of achieving
a stochastic design with an exponential distribution
of SOAs, which can buy you efficiency for the main
effect versus baseline at short SOA.
(You could model them nonetheless, at the cost of
an extra df in your model - they should have a
parameter estimate that doesn't differ significantly
from 0. One situation where this might not be true
would be if the null events were rare, eg <30% of
trials, because in this case they ARE likely to be
"detectable" - as "missing stimuli" in an otherwise
predictable sequence - and have a corresponding
impulse brain response at a specific point in time).
If your "null events" are detectable by subjects (eg
a change in the fixation cross), then they SHOULD
be modelled, but they are now really LOW-LEVEL
CONTROL events (they are no longer true "null events",
as in the original definition).
When they are not modelled, they simply comprise
part of the baseline and a [1] contrast on an event-type
of interest will compare that event-type to baseline.
When they are modelled (eg as a second event-type),
a [1 -1] contrast should give you similar results, but a
a [1] contrast will have less power (at short SOA) because
of the collinearity between your two columns.
If this is not clear, we will have to chat at SPM/HBM
in Japan!
Rik
"Scott H. Johnson" wrote:
> Should null events that are used to induce temporal variation in the onsets of various experimental conditions be explicitly coded in one's glm?
>
> I have been going back through past communications re: this topic and the issue seems unclear.
>
> Can someone articulate how spm handles implicitly vs. explicity coded vars?
>
> Thanks in advance
>
> Scott
--
---------------------------------------------------------
DR RICHARD HENSON
Wellcome Department of Imaging Neuroscience
& Institute of Cognitive Neuroscience
Institute of Neurology
University College London
17 Queen Square
London, WC1N 3AR
England
EMAIL: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~rhenson
TEL1 +44 (0)20 7679 1131
TEL2 +44 (0)20 7833 7472
FAX1 +44 (0)20 7813 1420
FAX2 +44 (0)20 7813 2835
MOB +44 (0)794 1377 345
---------------------------------------------------------
--
|