Prev: Triu function
Next: mnrfit
From: Beatrice Pöschel on 11 Aug 2010 12:49
hello
i know that increasing window size is increasing frequency resolution and that increasing the number of windows (i.e. small window size) helps to find low coherence values (because of increased averaging).
i have 2 question concerning window size:
1. i am wondering, if there is a limited window size if i am looking for coherence in certain frequency bands. i.e. theta vs gamma.
2. with different window sizes i would expect to get coherence spectra that have a different frequency resolution but are similar, e.g. have a peak at 10 hz. nevertheless i am getting coherence spectra with peaks in certain frequency bands and these peaks are not similar across coherence spectra with different window size. i dont understand why i get different results for different window size. do you have an idea where the error is?
thanks a lot!
best, beatrice
From: Wayne King on 11 Aug 2010 13:09
"Beatrice Pöschel" <beatrice.poeschel(a)gmx.de> wrote in message <i3uka1$bhn$1(a)fred.mathworks.com>...
> hello
> i know that increasing window size is increasing frequency resolution and that increasing the number of windows (i.e. small window size) helps to find low coherence values (because of increased averaging).
> i have 2 question concerning window size:
> 1. i am wondering, if there is a limited window size if i am looking for coherence in certain frequency bands. i.e. theta vs gamma.
> 2. with different window sizes i would expect to get coherence spectra that have a different frequency resolution but are similar, e.g. have a peak at 10 hz. nevertheless i am getting coherence spectra with peaks in certain frequency bands and these peaks are not similar across coherence spectra with different window size. i dont understand why i get different results for different window size. do you have an idea where the error is?
> thanks a lot!
> best, beatrice

Hi Beatrice, an important parameter that you need to consider is the sampling frequency here. In order to resolve peaks in the theta vs. gamma bands you just need to make sure that your window length is big enough given your sampling frequency.

As far as your second question, it's hard to answer without the specifics, but having a window size too short for an oscillation of a certain frequency would cause you to miss that oscillation, while lengthening the window would allow you to catch it. Also, by making the window size smaller, you can also have two separate peaks merging into one.

Wayne
 | 
Pages: 1
Prev: Triu function
Next: mnrfit