Dear SPMers,
in searching for a method to draw population-level inferences in a context
where a standard random-effects model does not seem adequate, I came across
the paper by Friston et al. (1999), "Multisubject fMRI studies and
conjunction analyses". My question is how to perform the kind of analysis
described in that paper using SPM8.
The key point of the paper is that there is a relation between the
significance levels on the single-subject level [alpha] and the population
level [alpha_c] with a bound [gamma_c] on the prevalence of the effect in the
population, given by Eq. 3. This equation can be used in two ways: Either
[alpha] and [alpha_c] are specified, and the result is [gamma_c], or [alpha]
and [gamma_c] are specified, and the result is a population-level empirical
significance level (p-value) with respect to the chosen bound.
In the example for a multisubject conjunction analysis given in the SPM8
manual, Sec. 31.3.10, the result (Fig. 31.15) gives the usual map and table
of p-values (on the set- and peak-level). This would imply that the analysis
uses a specific value of [gamma_c]. However, it does not seem to be possible
to specify it as a parameter to the analysis, nor could I find any
information on what fixed value might be built into it.
So I would like to know: Which value of [gamma_c] underlies the p-values
resulting from a multisubject conjunction analysis in SPM8, and how can I
specify my own choice of the bound (via GUI or command line)?
Thank you very much,
Best,
Carsten
PS: Please be aware that I'm asking specifically about the *multisubject*
conjunction, not a conjunction of different contrasts ("congruent" rather
than "incongruent" conjunction).
--
Carsten Allefeld, M.A., Dr. rer. nat.
Research Associate
Bernstein Center for Computational Neuroscience
Charité Universitätsmedizin Berlin
Philippstr. 13, Haus 6
10115 Berlin, Germany
Phone: +49 (0)30 2093 6766
Fax: +49 (0)321 2123 6854
|