Goodness-of-Fit Tests in Calibration: Are They Any Good for Selecting Least-Squares Weighting Formulas?

Abstract

Answer: No. Most goodness-of-fit (GOF) tests attempt to discern a preferred weighting using either absolute or relative errors in the back-calculated calibration x values. However, the former are predisposed to select constant weighting and the latter 1/x 2 or 1/y 2 weighting, no matter what the true weighting should be. Here, I use Monte Carlo simulations to quantify the flaws in GOF tests and show why they falsely prefer their predisposition weighting. The weighting problem is solved properly through variance function (VF) estimation from replicate data, conveniently separating this from the problem of selecting a response function (RF). Any weighting other than inverse-variance must give loss of precision in the RF parameters and in the estimates of unknowns x 0 . In particular, the widely used 1/x 2 weighting, if wrong, not only sacrifices precision but even worse, appears to give better precision at small x, leading to falsely optimistic estimates of detection and quantification limits. Realistic VFs typically become constant in the low-x, low-y limit. Thus, even when 1/x 2 weighting is correct at large signal, the neglect of the constant variance component at small signal again gives toosmall detection and quantification limits. VF estimation has been disparaged as too demanding of data. Why this is not true is demonstrated with Monte Carlo simulations that show only a few percent increase in calibration parameter uncertainties when the VF is estimated from just three replicates at each of six calibration x values. This point is further demonstrated using examples from the recent literature.

Approximated Uncertainty Propagation of Correlated Independent Variables Using the Ordinary Least Squares Estimator

“…The weighted SR matrix is then minimized by an iterative solver. When the uncertainties of independent variables and covariances among dependent and independent variables are nonexistent, the GLS regression can be reduced to the weighted least squares (WLS) regression [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 16 ]. As the off-diagonal elements of the SR matrix are zero, the analytical expression for the fitted parameters of the regression model can be analytically derived.…”

Linear Fitting of Time-Varying Signals in Static Noble Gas Mass Spectrometry Should Be Avoided

“…The goodness of a fitting in mathematics is determined by the variance structure of the data set. 13,14 Principally, many functions can be used to fit the experiment data with high R 2 . 3 With linear fitting, the output (viz.…”

“…However, in a few studies, the adopted fittings are parabolical, polynomial, or exponential. − In most of the application cases, linear function is widely used for the treatment of analytical responses to deduce the signal at T 0 on account of high R 2 with linear fitting. ,, Nevertheless, from the perspective of regression analysis, high R 2 is only a qualitative indicator to explain the distribution of a given data set, while it could not guarantee a right prediction or retrodiction result if it extrapolates to the full scale of unknown range. The goodness of a fitting in mathematics is determined by the variance structure of the data set. , Principally, many functions can be used to fit the experiment data with high R 2 . With linear fitting, the output (viz.…”