Gluhovsky, Ilya (2006, ASA/ASQ) Sun Microsystems Laboratories, Menlo Park, CA
[This abstract is based on the author's abstract.]Two separate approaches are considered for modeling the regression relationship between a predictor vector and a response. The first approach initially fits an unconstrained additive model, and then imposes the isotonic constraint to cause the least perturbation of the fit. The second approach optimizes the fitting criterion that includes a roughness penalty over isotonic smoothing splines. A theory is developed to incorporate a variety of smoothing methods as building blocks of the unconstrained model. It is shown that fitting an isotonic smoothing spline model produces superior results over a broad range of target functions. The methodology is applied to modeling computer cache miss rates.
Data smoothing,Isotonic regression,Model discrimination,Multiple regression,Multivariate,Nonparametric methods,Selection,Spline functions