The two basic components of a statistical model, the deterministic part and the stochastic part, are well separated in the penalized likelihood score L(f) + (λ/2)J(f) of (1.3). The deterministic part is specified via J(f), which defines the notion of smoothness for functions on the domain X. The stochastic part is characterized by L(f),which reflects the sampling structure of the data.
KeywordsHilbert Space Tensor Product Model Construction Reproduce Kernel Hilbert Space Smoothing Spline
Unable to display preview. Download preview PDF.