Volume 477, Number 3, January III 2008
|Page(s)||967 - 977|
|Published online||12 November 2007|
Advanced fit technique for astrophysical spectra
Approach insensitive to a large fraction of outliers
Faculty of Physics, University of Belgrade, Studentski Trg 12-16, Belgrade, Serbia e-mail: firstname.lastname@example.org
2 University of Nova Gorica, Vipavska cesta 13, Nova Gorica, Slovenia
Accepted: 23 September 2007
Aims.The purpose of this paper is to introduce a robust method of data fitting convenient for dealing with astrophysical spectra contaminated by a large fraction of outliers.
Methods.We base our approach on the suitable defined measure: the density of the least squares (DLS) that characterizes subsets of the whole data set. The best-fit parameters are obtained by the least-square method on a subset having the maximum value of DLS or, less formally, on the largest subset free of outliers.
Results.We give the FORTRAN90 source code of the subroutine that implements the DLS method. The efficiency of the DLS method is demonstrated on a few examples: estimation of continuum in the presence of spectral lines, estimation of spectral line parameters in the presence of outliers, and estimation of the thermodynamic temperature from the spectrum that is rich in spectral lines.
Conclusions.Comparison of the present results with the ones obtained with the widely used comprehensive multi-component fit yields agreement within error margins. Due to simplicity and robustness, the proposed approach could be the method of choice whenever outliers are present, or whenever unwelcome features of the spectrum are to be considered as formal outliers (e.g. spectral lines while estimating continuum).
Key words: methods: data analysis / methods: numerical / techniques: spectroscopic / line: profiles
© ESO, 2008
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.