Dear Anonymous,
If one intends on performing statistical inference at the time series level,
then there are two reasons why one should not use a standard ANOVA with BOLD
fMRI data. One is that there is temporal autocorrelation in BOLD fMRI time
series errors, which invalidates the specificity of inference in a standard
ANOVA. The second is that the signal (i.e., the systematic component of the
data) is by, physiological mechanisms, a temporally lagged and smoothed
version of neural activity, such that the correct (and hence the most
sensitive) regressor is not simply an dichotomous predictor but rather one
that manifests this temporal lag and smoothness.
I do not agree with the premise of your last question, as in principle you
can do a valid (from a specificity perspective) single-repetition (or single
cycle) BOLD fMRI experiment. The reason why one wouldn't if given the choice
is quite simply the more cycles, the more power for detection. This is of
course true for PET experiments as well, and my understanding is that it is
constraints on the exposure to radioactivity that limits the amount of data
per subject one acquires with PET.
Eric
----- Original Message -----
From: "apb 0628" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, July 08, 2005 2:40 PM
Subject: [SPM] PET model for fMRI ?
> Sorry if this seems a trivial question but I've only just started using
SPM
> to analyze some fMRI data having had some limited experience with PET data
> and I was wondering if some one could explain why you can't use a PET
model
> (e.g 10 scans EPI at rest 10 scans EPI active) to look for mean
differences
> in fMRI, why is it necessary to model it as a time series and have more
than
> one period of rest and activation.?
>
> _________________________________________________________________
> Is your PC infected? Get a FREE online computer virus scan from McAfeeŽ
> Security. http://clinic.mcafee.com/clinic/ibuy/campaign.asp?cid=3963
|
