COBS: qualitatively constrained smoothing via linear programming (Q1979097)

A huge amount of research has been carried out in the past few decades on nonparametric function estimation based on the idea of smoothing. A number of highly successful smoothing methods are available in \(S\)-plus. Among them are smoothing splines, kernel smoothing, local-span supersmoother, and robust smoothing. Several other smoothers are also available from \textbf{statlib}. Any smoother that performs local averaging over the response values will yield a fitted function falling within the range of the response values. In some applications, this is highly desirable. For example, if the response variable is age or income, zero is the intrinsic lower bound of the estimated function. Conventional spline methods, however, may not preserve this positivity near the boundaries. The constrained \(B\)-spline smoothing (COBS) method introduced in this paper can easily impose such boundary conditions and help overcome this weakness of spline smoothers. In the work by \textit{X. He} and \textit{P. Shi} [J. Am. Stat. Assoc. 93, No. 442, 643-650 (1998)] a special case is considered. It is shown that the \(L_{1}\) projection of a smooth function into the space of \(B\)-splines provides a monotone smoother that is flexible, efficient and achieves the optimal rate of convergence. Several options and generalizations are included in COBS: it can handle small or large data sets either with user interaction or full automation. Three examples are provided to show how COBS works in a variety of real-world applications.