Dear Experts,
I'd like continue the considerations about SLICE TIMING PROBLEM IN EVENT
RELATED FMRI begun some days ago with Rik and Vince.
I've an event related design with four interest regressors (events). I
model with canonical hrf (I don't perform slice timing correction, TR = 3
sec) and I follow this steps for each subject to conclude that maybe
canonical hrf doesn't fit well my data:
1) plot -> contrast estimates and 90% C.l. -> effects of interest -> the
betas are low;
2) plot -> fitted responses -> effects of interest -> adjusted -> scan or
time -> the fitted responses are lower and shifted forward in time than
adjusted data;
3) plot -> event-related responses -> "select an event" -> fitted response
and PSTH -> the PSTH has not canonical hrf shape because it shows two
peaks, the second peak is shifted forward regarding the first canonical
peak;
4) plot -> event-related responses -> "select an event" -> fitted response
and adjusted data -> the adjusted data are very spread around to the fitted
response.
On tha basis of this qualitative analysis and of the previous Rik advice I
think to precollect some pilot data and explore these parameters for
example performing a slice timing correction and/or modeling with an
informed basis set. If I find with this pilot that both canonical hrf betas
and time/dispersion derivative betas are significant I think to use a less
constrained basis functions set as a fourier set or a FIR model (Matthew
suggested me FIR model regarding fourier set because it is more simple to
understand).
Am I mistaking about plot understanding?
Thank you.
|