Dear Panagiotis,
As you already know using bandpass filter for such a wide band is bad
practice. It was in the tutorial chapter for historical reasons and I
have just changed it so in the next version of the manual it will not
appear any more. What happens is that Matlab needs to design bandpass
filter according to your specification and taking into account the
sampling rate of the data. For some combinations of sampling rates and
cut-offs the filter becomes unstable, you get NaNs at the output and
then you get that error. So only use bandpass filter for narrow bands.
Best,
Vladimir
On Fri, May 7, 2010 at 3:51 PM, Panagiotis Tsiatsis
<[log in to unmask]> wrote:
> Hello dear all,
>
> I am writing to you because I am facing a weird problem - after some
> preprocessing stages ( loading a continuus CTF AUX MEG data file, filtering
> the continuous file, downsampling the continuous file and then epoching) and
> although the whole procedure proceeds without errors, when I try to load
> back the data (via "Display" and then clicking on the MEG tab) I am getting
> nothing but an empty white window. When I try to select another trial, I am
> getting the following error:
>
> ??? Error using ==> set
> Bad property value found.
> Object Name : axes
> Property Name : 'YLim'
> Values must be increasing and non-NaN.
>
> Error in ==> spm_DisplayTimeSeries at 338
> set(ud.v.handles.gpa,...
>
> Error in ==> spm_eeg_review_callbacks>updateDisp at 994
> ud =
> spm_DisplayTimeSeries(D.data.y(:,:,trN(1))',options);
>
> Error in ==> spm_eeg_review_callbacks at 723
> updateDisp(D,1)
>
> ??? Error while evaluating uicontrol Callback
>
>
>
> After some playing around to see what is wrong and by leaving one
> preprocessing step out each time, is seems that the problem appears only
> when I apply Filtering and, weirdly enough, only when I apply a bandpass
> filter with a lower value of <= 1 Hz. The higher limit does not seem to
> affect anything. When I apply only a highpass filter with a cut off value of
> 1 everything works fine.
>
> I tried both Matlab 2009b and 2010a but the problem is always there. I also
> made a clean installation of SPM
>
>
> I have to also note that I have used similar scripts that utilized bandpass
> filtering without a problem in the past without a problem,on the same
> computer and same matlab / spm version
>
> Does anybody have a clue?
>
> PS: I filter the continuous file first and then I epoch so the epoch
> duration should not have anything to do with the filter bands that I am
> using
>
> PS2: I know that I could use a combination of a bandpass and a lowpass
> filter, I am just wondering why things are not working with the bandpass
>
|