Hi Erik,
This depends on the smoothing filter... You can see from the formulas in
http://www.cmrr.umn.edu/stimulate/frame/fwhm/img5.gif that when sigma
(FWHM) increases the Gaussian in the time domain becomes broader, but
the Gaussian in the frequency domain becomes more narrow. Only when
sigma==1 they are the same.
This means that when sigma>1 (smooth), and you use the same sigma in the
frequency domain, the frequency-domain version includes more frequencies
(closer to tha data).
When sigma<1 (less smooth) and you use the same sigma in the frequency
domain, the frequency-domain version includes fewer frequencies (further
from the data).
If you want your frequency filter to represent the right Gaussian
kernel, then the best way to that is to make the filter in the time
domain, and then compute the Fourier transform of that filter.
Best,
Alle Meije
Erik Chang wrote:
> Hi Alle,
>
> thanks for your comments. given these two data sets would be
> different, which of them is closer to the real data? and should the
> difference be huge?
>
> cheers,
> erik
--
Brain Mapping Unit
Dept. of Psychiatry
University of Cambridge
Downing Site
Cambridge CB2 3EB
tel: (+44) (0)1223 764676
fax: (+44) (0)1223 336581
[log in to unmask]
|