In the literature, there are contradictions regarding the
Nyquist frequency for irregularly sampled data. Some authors
(Press et al. 1992;
Horne et al. 1986)
identify it with
where
is the
"average'' sampling rate.
The Nyquist frequency is also often identified with
, where s is the smallest time interval in the
sample
(Scargle 1982;
Roberts et al. 1987).
Nevertheless,
Press et al. (1992) and
Roberts et al. (1987)
remarked that frequencies can be detected above their quoted
values.
For evenly sampled data, the Nyquist frequency is defined
as , where
is the time sampling step.
We will show how this quantity can be extended to the
irregular situation. In most cases, the Nyquist frequency is much
larger than usually thought. The present work is particularly
relevant to large databases that are becoming available (MACHO,
OGLE, EROS, HIPPARCOS, etc.).
Copyright The European Southern Observatory (ESO)