[Mne_analysis] Phase Locking Value and filtering

[Hari Bharadwaj]

Hi Elisabeth,
    Yes.. something like 150 Hz cutoff to analyze upto 100 Hz. Also, the results should be the same for unfiltered and filtered data for a far enough cutoff.

Regards,
Hari

Elisabeth Fonteneau <Elisabeth.Fonteneau at mrc-cbu.cam.ac.uk> wrote:

>Hi Hari,
>I am filtering the data when I am epoching with the function mne_epochs2mat
>As far as I understand your recommendation is to use a cut off at 150 low pass and to stop the analysis at 100 Hz for example?
>
>Thanks a lot again for your time
>Elisabeth
>
>
>
>-----Original Message-----
>From: Hari Bharadwaj [mailto:hari at nmr.mgh.harvard.edu] 
>Sent: 04 October 2010 15:13
>To: Elisabeth Fonteneau; mne_analysis at nmr.mgh.harvard.edu
>Subject: Re: [Mne_analysis] Phase Locking Value and filtering
>
>Hi Elisabeth,
>    It does sound like the PLVs you are seeing around the cutoff frequency of the filter are dominated by filtering ripples. Are you filtering the data before epoching? Using a good FIR filter before epoching with cutoffs far away from your frequencies of interest would probably be the ideal thing to do.  Since the frequency resolution of the wavelet transform becomes progressively poorer at higher frequencies, edge effects can bleed over a large range of frequencies around the cutoff.
>
>Regards,
>Hari
>
>Elisabeth Fonteneau <Elisabeth.Fonteneau at mrc-cbu.cam.ac.uk> wrote:
>
>>
>>Hello MNE users,
>>
>>First thank you for answering my previous questions about PLV.. that was very usefull...
>>I take the advantage to have some PLV expert to ask another related question about PLV & filtering (low pass)
>>
>>I am using the function provided by SPM (spm_eeg_morlet) with the following options:
>>Factor 7 and sampling rate 2 ms from 1 to 80 Hz to start with..
>>>From my results, I realised that some of my effects were starting in the gamma band (40-80Hz) but could extend in higher gamma range (the clusters were clearly cut at the 80Hz level).
>>
>>My original data were lowpass filter before the morlet transform to 80 Hz, I then use a different filter that time to a 150 Hz lowpass ... and recompute everything.
>>Now I was assuming that with my new data, I should find similar clusters within the 40-80hz gamma band and additional ones may be in the higher gamma band..
>>Indeed, by increasing my frequency range I found some new clusters in high gamma band, but my previous clusters in the lower gamma band completely disappear.
>>I am puzzle with this and when checking the PLV values, these are closed to be identical in the lower frequencies band (below 40Hz, but with lower value) but on the 40-80hz the PLV values have completely changed...
>>
>>I am thinking the problem is coming from the low pass filter that will induce some artefact at the edge of frequency band or may be the morlett transform properties?
>>So far I can not see any other explanation but you might have already one in your head, or may be you came across this problem too..
>>If my problem is coming from the filter, what will you suggest to compute an accurate PLV value, no filter at all? Or take a larger low pass filter that will not interfere with the frequency of interest?
>>If by chance you have a paper in mind that could direct me to the solution that will be very helpful
>>
>>Thanks for your answer
>>Elisabeth
>>
>>**********************
>>Dr Elisabeth Fonteneau
>>MRC Cognition and Brain Sciences Unit
>>15 Chaucer Road
>>Cambridge CB2 7EF
>>England
>>Tel:  +44 (0)1223 355294 (extension 263)
>>Fax:  +44 (0)1223 359062
>>
>>Email: elisabeth.fonteneau at mrc-cbu.cam.ac.uk
>>Web: http://www.mrc-cbu.cam.ac.uk/people/elisabeth.fonteneau/
>>**********************
>>
>>
>>_______________________________________________
>>Mne_analysis mailing list
>>Mne_analysis at nmr.mgh.harvard.edu
>>https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>
>