[Eeglablist] Morlet wavelet spectral bandwidth in newtimef( )

Pritchard, Arwel arwel.pritchard.18 at ucl.ac.uk
Wed Aug 21 03:25:26 PDT 2019


Hello again,

Further to my previous message, I've started using FieldTrip rather than EEGLAB for the time-frequency analysis. That seems to solve my issue.

Best wishes,

Arwel

________________________________
From: Pritchard, Arwel
Sent: 19 August 2019 14:41:07
To: eeglablist at sccn.ucsd.edu <eeglablist at sccn.ucsd.edu>
Subject: Morlet wavelet spectral bandwidth in newtimef( )


Hello,


I'm trying to run time-frequency decomposition with Morlet wavelets using newtimef(), and I can't work out how to specify the spectral bandwidth of the wavelets (or at least calculate what the bandwidth will be at any given frequency for the wavelets used by newtimef). I therefore seem to be in the dark about what the frequency resolution of my analyses will be at any given centre frequency.


I will give more specific information in case this is helpful: I'm new to both EEGLAB and time-frequency analysis. Before attempting to use newtimef, I calculated (based on my reading about wavelets) the parameters m and c of a mother wavelet (invariant across frequency bands) that would give the desired trade-off between spectral and temporal resolution.


The time window is m*SDt, where SDt is one standard deviation of the temporal envelope. I want this to be 500 ms at 4 Hz centre frequency, i.e. 2 cycles. So far so good - I can specify 2 cycles in newtimef(), or varwin = [2].


The spectral bandwidth should be m*SDf, where SDf is one standard deviation of the spectral bandwidth. SDf = f0/c, where f0 is centre frequency and c is a constant. All of the relevant inputs seem to relate to the time window, though. And although I know that m*SDt = 0.5 s when f0 = 4 Hz, I don't think I can solve for m or c with just this one  equation.


As I said, I'm new to this, so there may be something extremely obvious that I'm missing! Any advice would be much appreciated.


Many thanks,

Arwel Pritchard


More information about the eeglablist mailing list