Dear Chiau-Fang Chen,
|We'd like to analyze fMRI data from single trial design.
|The parameters are as follows: ISI=24secs, TR=3sec, and we choose hrf(only) as the basis function.
|Should we change the length of kernel from 32 to 24 when modeling?
No, you do not need to do this. The kernel represents the basis function
that is convolved with the delta functions you specify as trial
onsets within SPM to construct the design matrix. This basis function does
not need adjusting for the ISI.
Note that it is possible to use considerably shorter ISIs in event related
designs than the one you are using, as the haemodynamic responses to
individual trials are linearly separable, at least down to 1-2s ISI
(Friston et al 1998 MRM). Such a short ISI may be considerably more
powerful in estimating a differential event-related response between two
events (Henson & Josephs 1999 for further discussion).
best wishes,
G
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|