Least-squares Spline Regression with Block-diagonal Variance Matrices

Publisher Website
Google Scholar
Abstract
A numerical method of solution is presented for the least squares fitting of experimental data by spline functions in the case where the data errors are correlated and for which the variance matrix is specified. The method is general in that it permits (a) splines of any order, (b) the knots of the spline to be arbitrary in number and position, and (c) variance matrices that are block diagonal in form. Since limiting forms of (c) are diagonal and full variance matrices, the method can handle, as special cases, both conventional spline regression problems and spline regression problems with general, unstructured variance matrices. An application to gamma spectrometry, in which the blocks of the variance matrix have special structure, is fully treated.

This publication has 0 references indexed in Scilit: