I am comparing the efficacy of two filtering techniques on a simulated
time series that has random and systematic errors. As the data is
simulated, I know the frequencies and amplitudes that generate the
systematic noise. I'm looking for a way to compare the techniques in a
simulation framework - i.e., I will generate many instances of the time
series varying the parameters, perform the filtering, and test the
filters against the input. I have started doing this by combining the
filtered series and the systematic noise as a single time series using
ts.union() and then running spectrum() and comparing the coherencies of
the filters against the systematic noise at the right frequencies. This
seems to work but it is simplistic. Can anybody think of a tidy method
in R of comparing two time series against a known spectrum to test which
one preserves that spectrum better?
Best regards, Andy
