[vsnet-chat 3658] "n-sigma corrections"

[crawl@zoom.co.uk]

Dear Dr Kato

re "n-sigma corrections"

Sorry to show my ignorance, but these worry me...

...for example, AF Cygni.

To _show_ that AF Cygni has some sort of meaningful change in the long term
periodicity, the "eras" of periodicity need to be defined, and the data
split into appropriate subsets.  Each subset then undergoes fourier analysis.

To show that any different periods within the subsets are "real" and
distinct, the error range of the periods, and the significance of the
period peaks in the fourier spectrum have to be calculated [if the error
ranges of two eras of periodicities overlap, for instance, this may just be
random drift, whilst if they do not we can have some confidence of separate
and distinct eras of different periodicity].

This is problematic and non-trivial, as they say, because there is little
agreement between experts as to the valid routes for calculating these
quantities, but they do all tend to rely upon the use of the standard
deviation and/or noise of the raw light curve [minus the peak(s) in question].

In which light you'd have to be very, very certain that some sort of
n-sigma correction was _only_ removing observer problems, and not real
noise[?????]  I concede that it could be possible that the effect is trivial.


As regards intranight or intraweek erratic flaring in LPVs though, n-sigma
corrections sound a good way to miss these!  If they do exist at all...
...anynobody know the latest word on this apparent phenomena?

Cheers

John

JG, UK

VSNET Home Page

Return to Daisaku Nogami

---

vsnet-adm@kusastro.kyoto-u.ac.jp