Flexible parsimonious smoothing and additive modeling. (English) Zbl 0672.65119

The authors develop an approach to regression fitting based by a piecewise linear fitting. A procedure for deciding how many knots to use is described. The method is compared with other competitive techniques and it seems that it is better for small data sets in presence of high noise.
Reviewer: I.Grosu


65C99 Probabilistic methods, stochastic differential equations
65D10 Numerical smoothing, curve fitting
62J02 General nonlinear regression
Full Text: DOI