Rounding and uncertainties in parameters determined from fits to experimental data, or a failure to round data-analysis fit parameters properly may make them useless

Le Roy, Robert

Permalink

https://hdl.handle.net/2142/59142 Copy

Description

Title

Rounding and uncertainties in parameters determined from fits to experimental data, or a failure to round data-analysis fit parameters properly may make them useless

Author(s)

Le Roy, Robert

Issue Date

2014-06-16

Keyword(s)

Small molecules

Abstract

Almost no physically interesting physico/chemical parameter is determined directly from a measurement. Rather, they are determined by performing a least-squares fit of some model to a set of data. Unfortunately, there seems to be no commonly accepted set of `best practices' for determining how to round off such fitted parameter values to a minimum number of significant digits while ensuring that they retain the ability to reproduce the experimental data within their uncertainties. This sometimes results in lists of fitted parameters with no quoted uncertainties that have 2--3 times as many significant digits as the data being fitted, or to the results of an analysis being defined by parameters that are (unnecessarily?) quoted to more digits than normal computer double precision, which makes those results difficult or impossible to apply. Alternatively, it may also lead to fitted parameters being `over-rounded' so that the model no longer accurately represents the experimental data. This presentation describes a `best practice' to address these problems, offers a general-purpose least-squares fitting program that applies it, and provides an illustrative application of this approach in a study of the $A\,^1\Sigma_u^+ - X\,^1\Sigma_g^+$ system of Mg$_2$.

Publisher

International Symposium on Molecular Spectroscopy

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/59142

DOI

https://doi.org/10.15278/isms.2014.MG01

Copyright and License Information

Copyright 2014 by the authors. Licensed under a Creative Commons Attribution 4.0 International License. http://creativecommons.org/licenses/by/4.0/

Owning Collections