V.V.Vityazev
SPbU
It is generally believed that the classical Schuster periodogram must not
be used in the problem of finding the periodicities in the irregularly
spaced data. This follows from the fact that if the data is pure
noise, the statistical distribution of the Schuster periodogram is no
longer exponential. From this point of view, the so-called LS-spectra
based on the least squares fitting of a sine function to the data are
recommended. Nevertheless, in many situations the Schuster
periodograms and the LS-spectra are close to being identical. The
paper presents a comparative study of the Schuster periodogram and
the LS-spectra with respect to their statistical properties. It is shown
that the
likeness of the periodograms under consideration depends on the
properties of the spectral window corresponding to the
distribution of time points. The main results are:
a) all the estimators evaluated at frequency
are
identical if
;
b) the Schuster periodogram differs from the LS-spectra
at the frequency
, where
is the
frequency at which the spectral window has a large side peak due to
irregular distribution of time points;
c) the analytical
expression for the probability distribution of the Schuster
periodogram when the time series is assumed to be unevenly spaced pure
noise is found. It is shown that the probability distribution deviates
from the exponential law only at the frequencies
that
satisfy the condition
. The paterns of the
time points yielding such pathology are given.
The numerical examples for several situations typical in astronomy
illustrate these conclusions.