Dear Matthijs,
> 1. Is it possible to use the t-maps acquired from individual analysis
> in a group-analysis? That way, the group-analysis may be speeded
> up a bit.
For a second stage (random effects) group analysis, you would use the
contrast images (con*.img) not the t-maps (spmT*.img). For a first stage
(fixed effects) group analysis, you would generally estimate a design matrix
that combined all subjects/sessions. So in either case, you wouldn't use
t-maps from individual subjects in a group analysis.
> 2. We have performed an event-related experiment with a stimulus
> interval of 3 secs. What basis set should we use to estimate the
> hrf?
The choice of basis set depends on what hypothesis you have about the nature
of the haemodynamic responses, so in one sense it's entirely up to you. The
more complex basis functions (e.g. Fourier set) the individual components
are not physiologically meaningful so you will need to use F-contrasts for
statistical inference.
In general if you're just starting off I would use hrf alone as this will be
the simplest to interpret. Remember you will need to consider slice timing
issues in an event related design, particularly if you use such a
constrained basis set.
> 3. Should we select the option 'Interactions among trials (Volterra)?
No, not unless you want to explicitly model non-linear interactions among
trial types (see Friston et al 1998 MRM 39, 41-52 for an example of this).
The SPM manual (http://www.fil.ion.ucl.ac.uk/spm/course/manual/model.htm)
provides some more information about the options available in specifying a
statistical model and how to choose them.
Best wishes,
Geraint
--
Geraint Rees | Institute of Cognitive Neuroscience
[log in to unmask] | University College London
+44-(0)20-7679-5496 (work) | Alexandra House
+44-(0)20-7813-2835 (fax) | 17 Queen Square
www.fil.ion.ucl.ac.uk/~grees | London WC1N 3AR
|