[FieldTrip] optimum number of cycles

["Jörn M. Horschig"]

Hi Rodolphe,

I don't think that your questions have a 'best' answer, but see below 
for my opinion.

> In the fixed-widow example, using a hanning taper, it is said that a 
> fixed widow of 500 ms can estimate power of different frequency bands 
> like 2 Hz, 4Hz, etc...
> However, for 2 Hz it means only 1 cycle.
>
> In the frequency-dependent window length, 7 cycles per frequency bands 
> are used.
>
> My questions are, is it ok to estimate a power with one cycle?

I would say no, but it depends on what trade-off between precision in 
time and frequency space you want to achieve. Of course it can be done, 
but the estimate is much more unreliable than when taking more cycles.

>  What could be the optimum number of cycles?

It depends on your question and frequency band. I would use at least 
three to five cycles for the most prominent frequencies(theta, alpha, 
beta), maybe two for very low frequencies (delta and below) and at least 
six or seven for higher frequencies (>40Hz). Some people use a fixed 
window length for the gamma-range and frequency smoothing around 10Hz, 
independent of the exact frequency (see e.g. the Neuron paper by Siegel 
et al., 2008, they use 250ms and 12Hz smoothing)

> Is ot ok to compare the power of different frequency bands that have 
> been estimated using a different number of cycles?

Yes and no, because different frequency bands have different 
characteristics, and are influenced differently by the 1/f noise. Taking 
more cycles for a higher frequency band and less for lower sounds 
logical, given that a signal low SNR requires more smoothing to reliably 
estimate the frequency content. However, frequencies bands are due to 
this 1/f noise different in power, e.g. alpha band activity will always 
have more power than gamma band activity.

>
> Thanks a lot,

You're welcome.

Best,
Jörn

-- 
Jörn M. Horschig
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Neuronal Oscillations Group

P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

Contact:
E-Mail: jm.horschig at donders.ru.nl
Tel:    +31-(0)24-36-68493
Web: http://www.ru.nl/donders

Visiting address:
Trigon, room 2.30
Kapittelweg 29
NL-6525 EN Nijmegen
The Netherlands