Building Models With Smoothing Spline ANOVA Decompositions

                Yudong Wang

SS ANOVA decompositions are a general technique for building
multivariate nonparametric regression models. They construct
functional spaces with certain modular structures that parallel
the classical analysis of variance decompositions. Components in a
SS ANOVA decomposition have the same interpretations as in a
classical ANOVA model as main effects and interactions. Instead of
presenting the theory behind these decompositions, I will
illustrate how to construct SS ANOVA decompositions for fixed,
mixed and nested designs using a simple example.