> From [log in to unmask] Mon Jul 19 11:28:30 1999
> Original-Sender: [log in to unmask]
> Date: Mon, 19 Jul 1999 12:21:12 +0100
> Organization: Neurologische Klinik Universtaetsspital Zuerich
> X-Accept-Language: en
> References: <[log in to unmask]>
Dear Bruno,
> we have a question about the possibility to explore correlations of
> signal intensity with a performance variable in a block design fMRI
> study.
>
> The study is designed as follows: 4 conditions (ABCD) with 4
> repetitions each summing up to ABCDABCDABCDABCD (identical for all
> subjects), fixed intervals, 6 scans (4s) per condition. Furthermore,
> we have one performance value for the whole session for each subject.
>
> We are now interested how to explore areas that show correlated
> activity between the conditions and particularly how to find out which
> areas do so in relation to the performance variable.
To examine the relationship between activation and performance you are
implicitly looking for a condition x perfomance interaction. This
would best be done in a second-level analysis: Simply analyse all your
subjects using a subject-separable model (i.e. n sessions) and then
specify n contrasts looking at a specifc condition effect (e.g. A vs
B). Take the n contrast images and re-enter these into SPM using the n
performance scans as a covariate.
The equivalent fixed-effect analysis would not be so appropriate
because the different levels of the performance factor apply to
different subjects.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|