Martin Brown shared this with us in microsoft.public.excel.charting:

bj wrote:

have you tried generating a sine wave with approximate amplitude
and period using your X data. squaring the difference between the
real data and the calculated data, and totaling using solver to
then minimize the total of he sum of squares. by changing your
period and amplitude. You may have to play with the settings of
Solver and your initial period and amplitude numbers to keep it
from diverging before it converges.

You can certainly do that, although for a dataset with a known
periodic repetition it may be better to fold the data over the known
period of 24 hours. Lomb periodogram and similar methods implement
this approach. The full method can cope with gaps in the data too.

It is a lot easier if you know you can fold the data on a fixed
period like 24 hours. This also has the advantage that the periodic
waveform need not be a sine wave - its a technique popular with
variable star observers too. You may even find spreadsheets around to
do it.

Hope this helps,
Martin Brown

Martin,

I googled up Lomb periodogram and my first reaction is:

<keanu
Whoa!
</keanu


My second reaction: I think I have to go back to uni and follow a
course about all those interesting numbercrunching statistical analysis
methods. I wished I had paid more attention when fast Fourier
transformation was explained...

--
Amedee Van Gasse