I want to analyze 36 months of historical sales figures to see if I can
identify any time-series that will allow me to project them into the future.

My rusty math memory tells me that I can use Fourier Analysis to find out
the underlying frequencies and amplitudes. The Fourier Analysis add-in - plus
what I have read on the web - tells me that the number of data points that I
use must be a power of 2, so my choices are to use 2, 4, 8, 16, or 32 months
of data (I use 32). I have also learned that the frequencies reported by the
Add-in are listed in ascending multiples of 1/(32 months). From this I can
calculate the value of each frequency returned by the Add-in.

What I don't yet know how to do with the output from Fourier Analysis is:
1. How do I interpret the number reported for each frequency? I assume that
each one is a measure of the amplitude of the reported frequency, but I do
not know how to handle the 'real' and 'imaginary' components of the number.
At first blush, I think I would ignore the imaginary, but ...
2. Once I have the frequencies and the amplitudes of each component
waveform, I assume that I select to use only the more significant (higher
amplitude) frequencies. Help on this selection process will be appreciated.
3. To make my projection, how do I make sure that I am applying the
waveforms at the correct part of their cycle. For example, I could start all
of the waveforms at t=0, but that is not necessarily where the cycle should
start.

Overall, it seems that I am looking for the equation for the full waveform
that Fourier Analysis is calculating; so that if I insert the output of the
FA into it, I will get the equation for the projected waveform.

Thanks for your help.

Mike