[AstroPy] Lomb-Scargle False-alarm probability

Jacob Vanderplas jakevdp at cs.washington.edu
Mon Jun 27 17:02:37 EDT 2016

Hi Élie,
There's no built-in code to compute the False Alarm Probability. I left
that out mainly because I don't trust the FAP as it's normally used in
Lomb-Scargle analyses. The simplest FAP (i.e. z-score) that often appears
in the literature is actually only applicable if you choose a frequency at
random. Once you ask for the FAP of the peak of the periodogram, the
assumptions behind the simple z-score are no longer met (it's essentially a
case of inadvertent p-hacking). This multiple-hypothesis issue is partially
addressed in, e.g. Baluev (2007), but even that analysis ignores the fact
that in practice, heights of different peaks are correlated in a
non-trivial way, related to the fingerprint the survey window leaves on the
periodogram. In the end, I don't believe there is any analytic expression
that captures correct FAPs the way astronomers would generally like to use
them. The only way I've encountered to do this robustly is via Monte
Carlo/bootstrap sampling; if you want ideas on that, we discuss it in
chapter 10 of our book (Ivezic 2014).
Hope that helps,

 Jake VanderPlas
 Senior Data Science Fellow
 Director of Research in Physical Sciences
 University of Washington eScience Institute

On Mon, Jun 27, 2016 at 12:27 PM, Élie Bouffard <
Elie.Bouffard at usherbrooke.ca> wrote:

> Hi everyone,
> I'm presently working with the newly included Lomb-Scargle periodogram in
> Astropy. It is very good, but I have a question. Is it possible to compute
> the false-alarm probability? If so, how exactly?
> Thanks!
> -Élie
> _______________________________________________
> AstroPy mailing list
> AstroPy at scipy.org
> https://mail.scipy.org/mailman/listinfo/astropy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/astropy/attachments/20160627/d6d1ccaf/attachment.html>

More information about the AstroPy mailing list