Dear Afraim,
> If I select a windowed fourier set to model event-related responses (as
> opposed to a non-windowed). Will my f statistic be higher where the response
> within the time series resembles a canonical HRF (i.e. clear departure and
> return to baseline).
If you use a windowed fourier set, you'll constrain the model such that
you fit something in the time-frame, which starts off from zero and goes
to zero at the end. I don't think that in your described case your
F-statistic will be higher. This would only be the case, if your model
captured some additional variance in the data than another model with
degrees of freedom staying constant.
>
> When therefore would it be preferable to use a non-windowed set?
> Perhaps for 'better capturing responses already in mid air' say where stimulus
> onset preceded scanning. Is this the case?
The non-windowed Fourier set is good for capturing any signal in the
specified time-frame and within the frequency range of the Fourier set,
whereas a windowed Fourier set implements the constraint that you are
looking for a signal, which behaves as described above.
>
> Also, whereas my event-related responses modelled using a windowed set
> and plotted in spm, have tapered error bars and finish at zero they don't
> necessarily start at zero (0.25 for instance). What exactly does the 'hanning
> filtering' do then?
The Hanning windowing implements the modulation of all Fourier set basis
function with a function based on a cosine. The modulation looks like a
bell-shaped function.
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|