toffeebean
Newbie level 1
Hi all!
I applied a fourier transform to my data to create a power spectrum, and the spectrum has a shape that somewhat resembles red noise with strong low frequencies and attenuated high frequencies. How do I tell statistically which peaks are significant and which are not? I suppose I need to somehow put confidence bounds on the power spectrum- I thought about trying to fit a red noise curve to it to use that as the "null hypothesis" but I'm having a hard time (see attached diagram - my red noise has a lower amplitude trend than the power spectrum). Other ways I've though of are the Chi-test and F-test. If you could tell me more about autocorrelation and how to use that in a model for red noise, that would be great too.
Thanks!
Becks
I applied a fourier transform to my data to create a power spectrum, and the spectrum has a shape that somewhat resembles red noise with strong low frequencies and attenuated high frequencies. How do I tell statistically which peaks are significant and which are not? I suppose I need to somehow put confidence bounds on the power spectrum- I thought about trying to fit a red noise curve to it to use that as the "null hypothesis" but I'm having a hard time (see attached diagram - my red noise has a lower amplitude trend than the power spectrum). Other ways I've though of are the Chi-test and F-test. If you could tell me more about autocorrelation and how to use that in a model for red noise, that would be great too.
Thanks!
Becks