Cargando…

Undesirable effects of covariance matrix techniques for error analysis

Regression with $\chi^2$ constructed from the covariance matrix should not be used for some combinations of covariance matrices and fitting functions. Using the technique for unsuitable combinations can amplify systematic errors. This amplification is uncontrolled, and can produce arbitrarily inaccu...

Descripción completa

Detalles Bibliográficos
Autor principal: Seibert, David
Lenguaje:eng
Publicado: 1993
Materias:
Acceso en línea:https://dx.doi.org/10.1103/PhysRevD.49.6240
http://cds.cern.ch/record/249291
_version_ 1780885433401475072
author Seibert, David
author_facet Seibert, David
author_sort Seibert, David
collection CERN
description Regression with $\chi^2$ constructed from the covariance matrix should not be used for some combinations of covariance matrices and fitting functions. Using the technique for unsuitable combinations can amplify systematic errors. This amplification is uncontrolled, and can produce arbitrarily inaccurate results that might not be ruled out by a $\chi^2$ test. In addition, this technique can give incorrect (artificially small) errors for fit parameters. I give a test for this instability and a more robust (but computationally more intensive) method for fitting correlated data.
id cern-249291
institution Organización Europea para la Investigación Nuclear
language eng
publishDate 1993
record_format invenio
spelling cern-2492912023-03-12T06:04:04Zdoi:10.1103/PhysRevD.49.6240http://cds.cern.ch/record/249291engSeibert, DavidUndesirable effects of covariance matrix techniques for error analysisParticle Physics - LatticeMathematical Physics and MathematicsRegression with $\chi^2$ constructed from the covariance matrix should not be used for some combinations of covariance matrices and fitting functions. Using the technique for unsuitable combinations can amplify systematic errors. This amplification is uncontrolled, and can produce arbitrarily inaccurate results that might not be ruled out by a $\chi^2$ test. In addition, this technique can give incorrect (artificially small) errors for fit parameters. I give a test for this instability and a more robust (but computationally more intensive) method for fitting correlated data.A version of the covariance matrix technique for treating correlated data points can amplify systematic errors that are present in the data. In addition, this technique systematically underestimates errors. Both the amplification and the underestimate of the error are uncontrolled, and arbitrarily inaccurate results can be obtained. I discuss alternative procedures for handling correlated data.hep-lat/9305014CERN-TH-6892-93CERN-TH-6892-93oai:cds.cern.ch:2492911993-05-17
spellingShingle Particle Physics - Lattice
Mathematical Physics and Mathematics
Seibert, David
Undesirable effects of covariance matrix techniques for error analysis
title Undesirable effects of covariance matrix techniques for error analysis
title_full Undesirable effects of covariance matrix techniques for error analysis
title_fullStr Undesirable effects of covariance matrix techniques for error analysis
title_full_unstemmed Undesirable effects of covariance matrix techniques for error analysis
title_short Undesirable effects of covariance matrix techniques for error analysis
title_sort undesirable effects of covariance matrix techniques for error analysis
topic Particle Physics - Lattice
Mathematical Physics and Mathematics
url https://dx.doi.org/10.1103/PhysRevD.49.6240
http://cds.cern.ch/record/249291
work_keys_str_mv AT seibertdavid undesirableeffectsofcovariancematrixtechniquesforerroranalysis