> Dear Jing,
>
> For some combinations of sampling rates and cutoff the Butterworth
> filter may not be stable and that will generate NaNs. The latest SPM
> version should give an error in these cases but previously you could
> get NaNs as the output. In these cases you should try to either change
> your cutoff frequency or reduce the filter order.
>
> Best,
>
> Vladimir
>
> On Thu, Jun 16, 2011 at 10:27 AM, Jing Kan <[log in to unmask]> wrote:
>> Dear SPMer,
>>
>> Thanks for concern!
Dear Vladimir,
Thanks for your reply and explanation.
In my case, the MEG sampling frequency is 1000 Hz which I down-sampled as
800 Hz as first step. Then, I did the baseline correction. By now, the
data shows fine in the "display" function. Next, I use the bandpass :
(0.03Hz~ 190Hz) in the filter process. Before I start to epoch, I checked
the data with the "dispaly" again, it turned to shown nothing.
Following your suggestion, I changed the sampling frequency back to 1000
Hz, and altered the bandpass frequency with (0.05Hz, 150Hz) and (0.01Hz,
200Hz), respectively. However, it still shows the same results.
However, if I reverse the order, to do the "epoch" first, and "filter"
next, the data shows fine in the display window. Considering that
"filtering" the data before "epoch" can generate the accurate result, I
just want to make sure "filtering" is implemented first. But, in this
case, should I just compromise to do the "epoch" first, or you have any
brilliant suggestion about it?
Looking forward to you reply! Many thanks in advance!
Best,
Jing
>>
>> I've got a problem when I use the SPM for the MEG data preprocessing:
>>
>> With my data, if I "Epoch" first and "Filter" after, the data looks
>> fine,
>> and I can also visualize my data in time course or scalp by
>> "dispay"-->"M/EEG" ; however, if I reverse the order to do the "filter"
>> first with some band pass filter and then do "Epoch", the data generate
>> lots of "NaN" and I can see nothing of MEG data in "dispay"-->"M/EEG".
>> May
>> I ask what possibility this problem might be coursed?
>>
>> Looking forward to your reply, and thanks in advance!
>>
>> Best,
>>
>> Jing
>>
>
|