A CRITICAL COMPARISON OF LEAST ABSOLUTE DEVIATION FITTING (ROBUST) AND LEAST-SQUARES FITTING - THE IMPORTANCE OF ERROR DISTRIBUTIONS

被引：51

作者：

MATHESON, IBC

机构：

[1] On Line Instrument Systems, Jefferson, GA 30549, P.O. Box 111

来源：

COMPUTERS & CHEMISTRY | 1990年 / 14卷 / 01期

关键词：

D O I：

10.1016/0097-8485(90)80007-O

中图分类号：

O6 [化学];

学科分类号：

0703 ;

摘要：

Non-linear least absolute deviation fitting of data has been shown to be superior to least squares where the data errors are unevenly distributed about the function. The methods give insignificantly different results for evenly-distributed errors. Criteria are given for choosing between least absolute deviation and least squares error minimization. Optical density measurements have non-Gaussian error distributions and are better fitted with minimization of the absolute deviation. © 1990.

引用

页码：49 / 57

页数：9

共 10 条

[1]

BEVINGTON PR, 1969, DATA REDUCTION ERROR, P32

[2] ANALYSIS OF DATA FROM THE ANALYTICAL ULTRA-CENTRIFUGE BY NON-LINEAR LEAST-SQUARES TECHNIQUES [J].

JOHNSON, ML ;

CORREIA, JJ ;

YPHANTIS, DA ;

HALVORSON, HR .

BIOPHYSICAL JOURNAL, 1981, 36 (03) :575-588

[3]

JOHNSON ML, 1985, METHOD ENZYMOL, V117, P301

[4]

Knuth D. E., 1981, ART COMPUTER PROGRAM, V2, P116

[5]

Levenberg KA., 2018, Q APPL MATH, V2, P436, DOI [DOI 10.1090/QAM/10666, 10.1016/j.healun.2011.11.021, 10.1090/qam/10666]

[6] AN ALGORITHM FOR LEAST-SQUARES ESTIMATION OF NONLINEAR PARAMETERS [J].

MARQUARDT, DW .

JOURNAL OF THE SOCIETY FOR INDUSTRIAL AND APPLIED MATHEMATICS, 1963, 11 (02) :431-441

[7] THE NON-EQUIVALENCE OF PADE-LAPLACE AND NON-LINEAR LEAST-SQUARES DATA FITTING - A PADE-LAPLACE BIAS TOWARDS SLOWER PROCESSES [J].

MATHESON, IBC .

COMPUTERS & CHEMISTRY, 1989, 13 (04) :385-386

[8]

MATHESON IBC, 1987, ANAL INSTRUM, V16, P345

[9]

PRESS WH, 1986, NUMERICAL RECIPES AR, pCH7

[10]

[No title captured]

← 1 →