MathGroup Archive 2004

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: power spectrum

  • To: mathgroup at smc.vnet.net
  • Subject: [mg52764] Re: [mg52695] power spectrum
  • From: yehuda ben-shimol <benshimo at bgu.ac.il>
  • Date: Mon, 13 Dec 2004 04:21:57 -0500 (EST)
  • References: <200412100123.UAA18915@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

Power spectrum is defined as the Fourier transform of the 
autocorrelation function. For discrete time series use Mathematica's 
Fourier function (FFT implementation for the DFT) rather than the 
FourierTransform.
A common practice (although known to be a biased estimation) is to take 
the absolute value of the Fourier outcome, that is Abs[Fourier[data]].
If you look for some specific spectral peek for low frequency, it is 
recommended that the average value of the data will be subtracted from 
the data before the analysis.
good luck
yehuda


George Szpiro wrote:

>If I have a string of data - let us say a time series of 200 observations - 
>how can I find the power spectrum? I did Fourier[list] but have no idea what 
>the output means.
>
>Thank you for any help (please to my email address:
>george at netvision.net.il)
>
>George Szpiro
>
>  
>


  • References:
  • Prev by Date: Re: Matrix m x n
  • Next by Date: Re: Sorting question
  • Previous by thread: Re: power spectrum
  • Next by thread: Re: power spectrum