Numerical Comparisons of Several Algorithms for Treating Inconsistent Data in a Least-Squares Adjustment of the Fundamental Constants

摘要

A number of recently proposed algorithms for treating inconsistent or discrepant data in a least-squares adjustment of the fundamental physical constants, along with several new but related algorithms, are compared in detail. The comparisons are first made by means of the numerical results the algorithms yield when applied to the same data considered by Cohen and Taylor in their 1973 adjustment which led to the recommended set of constants adopted by CODATA and in current use. A selected number of the algorithms are then further compared through the numerical results they yield when applied to the data considered by Taylor, Parker and Langenberg in their 1969 adjustment and by Cohen and DuMond in their 1963 adjustment. The principal conclusion of this paper is that the actual algorithm used to carry out an adjustment is much less important than the data finally selected for inclusion in the adjustment.