Fitting a straight line when both variables are subject to error: pharmaceutical applications

In many pharmaceutical applications one postulates a linear relationship between variables. The usual linear least-squares methods are appropriate when the values of the independent variable are constants, and the dependent variable is subject to error. When both variables are subject to error, as i...

Full description

Saved in:
Bibliographic Details
Published inJournal of pharmaceutical and biomedical analysis Vol. 12; no. 10; p. 1265
Main Author Roy, T
Format Journal Article
LanguageEnglish
Published England 01.10.1994
Subjects
Chemical Phenomena
Chemistry, Pharmaceutical - methods
Chemistry, Physical
Least-Squares Analysis
Models, Statistical
Online AccessGet more information
ISSN0731-7085
DOI10.1016/0731-7085(94)00057-3

Cover

More Information
Summary:In many pharmaceutical applications one postulates a linear relationship between variables. The usual linear least-squares methods are appropriate when the values of the independent variable are constants, and the dependent variable is subject to error. When both variables are subject to error, as in assay validation, calibration, and general correlation, the measurement error model (also called errors-in-variables) should be used especially when independent variable error is appreciable. In this paper, the theoretical properties of errors-in-variables methods are demonstrated with examples, and a technique for assessing the variability of parameter estimates without normality assumptions is presented. Robust methods resistant to outliers and not requiring normality assumptions, are also described.
ISSN:0731-7085
DOI:10.1016/0731-7085(94)00057-3