| Título: | Fitting a linear functional relationship to data with error on both variables |
| Autores: | THOMPSON, MICHAEL |
| Tipo de documento: | texto impreso |
| Editorial: | Cambridge [REINO UNIDO] : Royal Society of Chemistry, 2002 |
| Colección: | AMC Technical Brief, num. 10 |
| ISBN/ISSN/DL: | 29716 |
| Dimensiones: | 3 p. |
| Nota general: | ISBN: |
| Langues: | Inglés |
| Clasificación: | |
| Resumen: | It is fairly well known that a basic assumption of regression is that the y-values (the dependent- or response variable) are random variables while the x-values (the independent- or predictor variable) should be error-free. This model often prevails (or is approximated to) in analytical chemistry applications, for example in many calibrations. If the condition is violated, however, the results of the regression are in principle incorrect and in practice can be sufficiently incorrect to be misleading. What is far less well known is that a general method, functional relationship estimation by maximum likelihood (FREML), is available for use when the regression assumption is incorrect. FREML provides estimates of the intercept (a) and slope (b) of the line, plus their standard errors, that do not suffer from the biases introduced by the inappropriate use of regression. The method is symmetric in that the x- and y-variables can be interchanged without affecting the outcome. It is capable of handling heteroscedastic data, that is, data points with different precisions. |
Documentos electrónicos (1)
 27966 Adobe Acrobat PDF |
