Page 1 of 1

EW error and statistics

Posted: Thu May 17, 2018 8:06 pm
by Peter Somogyi
Hello,

I'm opening this topic off-branched from here:
http://www.spectro-aras.com/forum/viewt ... 100#p10767
I have no problem removing my comments being offtopic in that thread back to this link date, restoring its intended state, while I can save here the important thoughts.
Observers who just want to observe, should skip all of this new topic's discussion.

Firstly, I'd like to say many thanks to Benjamin M. who brought these articles to my attention:
May 14, 2018 Hα spectral monitoring of Aurigae 2009-2011 eclipse B. Mauclaire1 , C. Buil2 , T. Garrel3 , R. Leadbeater4 , A. Lopez5

From here, the relevant reference to this topic is:
Chalabaev, A. and Maillard, J.P. 1983, A&A, 127, 279

and an enhancement (also the more readable and better focused to me):
K. Vollmann and T. Eversberg, 2006, Remarks on statistical errors in equivalent widths

From the comments, I could read that SpcAudace already have implemented the Chabalaev approach:
http://spcaudace.free.fr/news/opus/#ew

though I don't quite understand how the spectrum serie fits into the example of " spc_ew del_sco_20100429.fit 6550 6578 4 n" - I can see here only 1 spectrum, whilst the article above seem to me using the whole serie of a nearby continuum to determine sigma.

Apart from EW measures, to correlate the EW result serie of different features, I've got the idea to use a correlation function from GDL (http://gnudatalanguage.sourceforge.net/) which is a free (and limited) implementation of IDL. It works to some extent with an astronomical package (https://idlastro.gsfc.nasa.gov), though importing fits had to google for another solution (ask me when interested). Nice tool to remain in the problem space (e.g. easily work on vectors, matrices) whilst allows some programming (again: in the problem space, no tcl no bash no pyton/C++ etc...). Fully free, and you don't need it if already have matlab. Just mentioning as a nice convenicence tool, this GDL has many alternatives (and resembles to as a very limited subset of matlab).

Others having ideas how to calculate EW error of a spectrum serie with the available free tools (IRAF, Audace, VSpec, ...) are welcome here. So far SpcAudace mentioned, yet to clarify which formula is used exactly and how to give a whole spectrum serie as a parameter.

Hope to read further thoughts and insights and practices here.

Peter