fft for financial time series

Hi

I'm trying to apply fft() to the spectrum analysis of a financial time series with 2608 (10 years) data with daily frequency... Can anyone suggest me how can I calculate freq and Pyy vectors from fft()?

Regards Marco

Answer 1

[Wayne King]

Below is an example.

   t = 0:2607;
   % just an example
   % put a sine wave with a period of 120 days and add noise
   x = cos(2*pi/120*t)+randn(size(t));
   xdft = fft(x);
   % because the length is even take the first N/2+1 points
   xdft = xdft(1:length(x)/2+1);
   Pyy = 1/length(x)*abs(xdft).^2;
   % create frequency vector
   df = 1/2608;
   % frequencies in cycles/day
   freq = 0:df:1/2;
   plot(freq,10*log10(Pyy));
   xlabel('cycles/day'); ylabel('dB/cycle/day');
   % find maximum value
   [maxval,index] = max(abs(xdft));
   % frequency corresponding to maximum value
   freq(index)

[Marco]

Hi Wayne King, thanks a lot for your help

bye, Marco

[Virgil Stokes]

How can you be certain that aliasing has not occurred?

[Bjorn Gustavsson]

...and why run an fft on financial data - that most certainly isn't even remotely stationary.