Abstract
THE methods of graduating or “smoothing ” a series of more or less irregular data affected by errors of observation or of sampling have of recent years received a good deal of attention; the present tract is a very valuable contribution to the subject. New methods are developed, which may be described as combinations of tangential or osculatory interpolation with the least-square method of Dr. Sheppard, and these and older methods are compared, both graphically and by finding the sum of the squares of the departures from the graduated curve. The results of this test are rather surprising. The osculating methods used give distinctly worse results than the other four methods employed. By grouping the observations it is possible to estimate what order of differences is negligible; if a higher order of differences is employed, the resultant curve tends to bring out something that is not inherent in the data, with very unsatisfactory consequences to the fit obtained.
Department of Applied Statistics (Computing Section), University of London, University College: Tracts for Computers.
Karl
Pearson
Edited by. No. 6, Smoothing. By E. C. Rhodes. Pp. 60+3 diagrams. (London: Cambridge University Press, 1921.) 3s. 9d. net.
Article PDF
Rights and permissions
About this article
Cite this article
Department of Applied Statistics (Computing Section), University of London, University College: Tracts for Computers . Nature 108, 495 (1921). https://doi.org/10.1038/108495b0
Issue Date:
DOI: https://doi.org/10.1038/108495b0