Dear Vladmir,
Can we rescale single trial TF before average?
Many thanks!
Bo-Cheng
引述 Vladimir Litvak <[log in to unmask]>:
> Dear Muhammad,
>
> The way one would usually do TF analysis is to apply TF to an epoched
> file, then average the result and then rescale. If you apply TF to an
> ERP which is quite noisy you'll get basically single trial TF which is
> noisy as well.
>
> Filtering will not help since you will either remove your frequency
> range of interest or not affect it.
>
> Best,
>
> Vladimir
>
>
> On 20 Feb 2011, at 03:51, MP <[log in to unmask]> wrote:
>
>> Hi all,
>>
>> I am doing the TF analysis on EEG data and I see some worrisome
>> "spots" in my rescaled TF, which translate as sharp "dips" in the
>> waveform of averaged TF over 12-20Hz when I convert TF into image.
>> I know that these spots could be from the noise (since it is a
>> single subject data), but I am worried that they will influence the
>> statistics.
>>
>> I have attached the ERP waveform (-200 to 2000ms), its TF
>> representation (4-20Hz, -198-2000ms), rescaled TF (using Log
>> rescaling, baseline=-148 to 0), and the waveform when TF is
>> averaged over 12-20Hz range.
>>
>> Is it possible that since these dips are random they will not
>> influence the results in the scalp-space considerably? Is there
>> anything I can do to minimize these dips, like lowpass filter the
>> ERP before doing TF?
>>
>> thanks for your help in advance
>>
>> - Muhammad
>> <Average_ERP_Cz.jpg>
>> <TF_Wavelet6_Cz.jpg>
>> <RescaleTF_Log_Cz.jpg>
>> <Average_12-20Hz_Cz.jpg>
>
|