Abstract
IN the course of some recent investigations in experimental physics we have processed a great deal of data with a digital computer using a variety of weighted least-squares fitting programmes. All these programmes, regardless of the specific problem, have the following feature in common. After the best values of the fitted parameters have been obtained a test of each data point is made. If the data point lies more than a specified interval from the best curve, the experimental and fitted values of the ordinate, and the associated uncertainty in the experimental value are punched out. We would like to emphasize the great value of this feature. As a criterion for punch-out a deviation equal to three standard errors was chosen. This corresponds to the well-known 99.3 per cent confidence limit, and eliminates nearly all statistical fluctuations from being punched. Within two months of adopting this procedure we had already experienced several resulting benefits.
Similar content being viewed by others
Article PDF
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
PAUL, D., SAWYER, W. Curve Fitting using a Digital Computer. Nature 201, 174 (1964). https://doi.org/10.1038/201174a0
Issue Date:
DOI: https://doi.org/10.1038/201174a0
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.