[vsnet-computing 14] (fwd) [vsnet-chat 4699] Re: period analysis

[Taichi Kato <tkato>]

(fwd) [vsnet-chat 4699] Re: period analysis

From: crawl@zoom.co.uk
Date: Sun, 05 Aug 2001 08:12:37 +0000
Subject: [vsnet-chat 4699] Re: period analysis

This probably doesn't apply here, but I'll mention it anyway.

If you're using more than one piece of software to examine PDM you'd better
check that the default settings are the same.

Different numbers of bins, covers, "fineness"/resolution step-sizes and
test period ranges can make a difference to pdm results in practice it
seems.  Probably because the method is quite prone to aliasing.

As I say, this may not matter here, because I think the vsnet sourced PDM
programs for both windoze and dos were written by the same author, so are
liable to have the same defaults [and bins and covers nearly always seems
to be 5 and 2 anyway], but if the data is a bit noisy this could be a concern.

As Dr Kato says, phase diagrams folded on a derived period value are a
quick and easy test, with an option for doing that easily even being built
into winpdm.

And as Dr Kato also says, the significant figure issue is a frequent
problem with Julian Dates, as most software tends to round up decimals
rather than clipping of the front end of the number.  As there appears to
be a tendency for people to accept output from machines as is, no matter
how many decimals the machines through out, you've to watch this.  I've had
people pass on to me PEP deltaV values to ten decimal places, because the
pepV calculating software was probably simply set to long integer precision.

Cheers

John

VSNET Home Page

---

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