Yn erthygl , sgrifennodd
Bernard Burton :
Alastair, you need to be a little cautious when trying to deduce trends or
cycles using running averages. It is a mathematical fact that any series of
unweighted means will show periodic fluctuations with a period two to three
times the averaging period. The effect has been discussed several times in
'Weather', (Burton June 1997, p195, Reynolds Weather 1978, pp 74-76).
Running averages of random numbers will produce some really convincing
cycles.

That's interesting - I didn't realise this. I was using running averages
a few years ago on data from a Raman spectrometer. But my method used
a sine wave average. So the number at the centre was given a weighting
of 1, the number at the edge (say 5 either side) 0, and the numbers
in between were cos((x/5)*90) - where x is the distance from the centre of
the averaged area. Multiply the result by the appropriate
number (I forget...) to rescale it into the same range as the original.

Using this method I never noticed any such cycles.

Adrian
--
Adrian Shaw ais@
Adran Cyfrifiadureg, Prifysgol Cymru, aber.
Aberystwyth, Ceredigion, Cymru ac.
http://users.aber.ac.uk/ais uk