Notice that since arch.dat is non-stationary, no time averaged
information is meaningful! Once can compute the numbers, but they
cannot characterize the properties of the data!
amplitude.dat
whatisit.dat
First zero of the auto-correlation function
8-9
10
exponential decay of envelope
data set too short
-0.0025
lag from 2-dim delay embedding
8
3
Theiler window
about 20
about 20
The Theiler window should be half a period for such an almost periodic
pattern, but is suitablt chosen much larger, since typically not much
statistics is lost, as long as it remains smaller than about 10% of
the data set size.