Hi All,
I apologize in advance if these questions seem too simplistic.
I am trying to find the best way of dealing with temporal
autocorrelations/serial correlations in my fMRI data. From the
literature I have been reading, I get the impression that temporal
autocorrelation and serial correlations in the data are the same thing.
Is this incorrect? If so, can someone clarify?
Because I think they are the same, the ReML approach that SPM5 takes
to account for the serial correlations and the use of the AR(1) to
model autocorrelations seems redundant to use when doing an SPM5
analysis of the data. Should both be used in an fMRI analysis to deal
with non-sphericity issues in the error terms? Is the use of both of
these enough to bring the error terms as close as possible to i.i.d.
form?
On a different topic, does anyone know where I can find example
regressors that are placed in the GLM to model effects of non-interest
such as biorhythms (cardiac/respiratory motion)?
Thank you.
Sincerely,
Todd Penney
|