[Eeglablist] choice of cycles for Morlet wavelets

[Celeste Reinking]

I'm familiar with the arguments for having a relatively high number of
cycles (over 5) when looking at high-frequency oscillations using Morlet
wavelets.  However, I don't know how to make a choice of cycles for low
frequencies (4 to 8 Hz, or 8 to 12 Hz.)  Having a lower number of cycles
would be desirable because it would improve the temporal resolution of the
frequency decomposition, and make it possible to look at low-frequency data
in relatively short epochs.   However, I have been told that I should also
consider the effects of the number of cycles on the standard deviation in
time and in frequency.   Since cycles = (central
frequency)/(sigma(frequency)), a 2-cycle wavelet with a central frequency
of 4 Hz would have a standard deviation in frequency of 2Hz.  And since
sigma(t) = 1/(2*pi*sigma(f)), this would give me a standard deviation in
time of 80 ms.  Assuming I were okay with the sigma(f), should I be
satisfied with the sigma(t) of 80 ms?  What criteria should I use in
determining whether a given sigma(t) is acceptable?  Or, more generally, is
there a minimum number of cycles needed to do an accurate wavelet
transform?
Thanks, Celeste Reinking