An alternative to R^2 to characterize the quality of fit for linear least squares using offsets of variable orientation related to uncertainties in the data.

26 May 2023, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

It is not appropriate to use the determination coefficient, R^2, to characterize the quality of fit for a least squares fitted line. In this paper, the maximum of R^2 is found as a function of the rotation angle of the data and gives the quality of fit for the line found by linear least squares with perpendicular offsets. The same rotation method is used to derive the perpendicular offset fit to the data, which yields two possible solutions where the correct root can be identified by a simple discriminant. These results are then generalized for any arbitrarily oriented offset, bringing about a new measure for the quality fit of a line, Q^2. Unlike the determination coefficient, R^2, this quality of fit measure is invariant to rotational transformations of the data and is specific to the offset’s orientation, which is directly related to the uncertainties in x- or y-data. Finally, this paper provides a method to determine the slope and intercept of a fitted line, as well as its quality of fit, given any estimate of the uncertainty ratio.

Keywords

quality of fit
linear least squares
perpendicular offsets
variable offsets
uncertainty
data fitting

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.
Comment number 1, Leslie Glasser: Sep 22, 2024, 01:42

I draw your attention to the following publications which relate to your work. Deming Regression is much used in clinical chemistry for method comparison. Wikipedia - Deming regression. https://en.wikipedia.org/wiki/Deming_regression (accessed December, 2023). Xu, S., A Property of Geometric Mean Regression. The American Statistician 2014, 68, 277-281. Cornbleet, P. J.; Gochman, N., When Linear Regression Gets Out of Line: Finding the Fix. Clin. Chem. 2020, 66, 1238-1239.