COBS: COnstrained BSplines
Abstract
COBS (COnstrained BSplines), written in R, creates constrained regression smoothing splines via linear programming and sparse matrices. The method has two important features: the number and location of knots for the spline fit are established using the likelihoodbased Akaike Information Criterion (rather than a heuristic procedure); and fits can be made for quantiles (e.g. 25% and 75% as well as the usual 50%) in the response variable, which is valuable when the scatter is asymmetrical or nonGaussian. This code is useful for, for example, estimating cluster ages when there is a wide spread in stellar ages at a chosen absorption, as a standard regression line does not give an effective measure of this relationship.
 Publication:

Astrophysics Source Code Library
 Pub Date:
 May 2015
 Bibcode:
 2015ascl.soft05010N
 Keywords:

 Software