Dear Jennifer,

See my comments inline:

> I have used Morlet wavelets to analyze preprocessed MEG data from the HCP in 
> resting state and the n-back working memory task. The input parameters that I 
> have selected are as follows:
> 
>  
> cfg = [];
> cfg.channel    = 'MEG';                
> cfg.method     = 'wavelet';               
> cfg.width      = 7;
> cfg.output     = 'pow';
> cfg.foi        = 4:0.1:100;                 
> cfg.toi        = -0.5:0.1:1.5;        
>  
> 
> My questions are:
> 
>  
> 1. Are these reasonable parameters for using the Morlet wavelets on the HCP 
> data? Can I use the same parameters for rest and working memory data? I would 
> like to be able to compare the results of working memory and resting analyses.
> 

No, these parameters are not really reasonable. Given some fundamental signal 
processing constraints (because we are dealing with finite and discretely 
sampled data) it is impossible to attain a frequency resolution of 0.1 Hz. In 
other words, it is not possible to distinguish frequencies at such a high 
resolution. The data traces (in your case the window width, which amounts to 7 
times an oscillation period at each frequency) are simply not long enough for 
that. Also, physiologically, the bandwidth of true rhythmic components in the 
electrophysiological data is on the order of half an octave to an octave (e.g. 
alpha starts at 8 and runs until 12, beta starts at 15 and runs until 25 Hz). 
There is no need to sample at such high frequency resolution. 
For a spectral analysis of the resting state data, I wouldn’t advice a 
time-frequency decomposition as you propose. In a task-based setting ‘time’ 
(relative to an experimental event) maps unto something externally meaningful, 
and in this context averaging the same time points across observations makes 
sense to increase your signal. In a resting state context ‘time’ is rather 
arbitrary (it does not map unto something externally meaningful) and averaging 
makes much less sense.

> 2. Is there something else I should do to the rMEG Preprocessed or 
> tMEGPreproc Preprocessed data before conducting frequency analysis?
> 
You’re safe to go if you use the rmeg/tmeg preproc data.

> 3. If I were to use Fourier analysis as a complementary method, can you 
> suggest the best window, i.e. Hanning taper? Multitapers? My initial 
> questions importantly depend on frequency resolution (would need less than 
> 1Hz resolution. currently set to 0.1). The default foi increment was 1 Hz, is 
> 0.1 valid and properly accepted by your scripts?
> 
See my comments about frequency resolution above. Although increments of 0.1 Hz 
are accepted by ft_freqanalysis, it does not really make sense. If you want to 
make a comparison between wavelets and the more flexible windowed FFT approach 
(which I would recommend to use) I suggest you have a look at the FieldTrip 
wiki: www.fieldtriptoolbox.org/tutorial/timefrequencyanalysis 

Best wishes,
Jan-Mathijs

> 
> Thanks very much, 
> 
> Jennifer
> McGill, Montreal Neurological Institute and Hospital
> 
> Department of Neurology and Neurosurgery; 3801 University st, MP116, 
> Montreal, QC H3A 2B4
> 
> 
> Website: jennifergoldman.webs.com
> 
> 
> _______________________________________________
> HCP-Users mailing list
> [email protected]
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
> 


_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Reply via email to