P-spline Smoothing for 2D Functional Data — pspline.2d • fdars

Fits 2D P-splines with anisotropic penalties in both directions.

Usage

pspline.2d(
  fdataobj,
  nbasis.s = 10,
  nbasis.t = 10,
  lambda.s = 1,
  lambda.t = 1,
  order = 2,
  lambda.select = FALSE,
  criterion = c("GCV", "AIC", "BIC")
)

Arguments

fdataobj: A 2D fdata object.
nbasis.s: Number of B-spline basis functions in s direction.
nbasis.t: Number of B-spline basis functions in t direction.
lambda.s: Smoothing parameter in s direction.
lambda.t: Smoothing parameter in t direction.
order: Order of the difference penalty (default 2).
lambda.select: Logical. If TRUE, select lambdas automatically.
criterion: Criterion for selection: "GCV", "AIC", or "BIC".

Value

A list of class "pspline.2d" similar to pspline().

Details

The 2D penalty uses Kronecker product structure: $$P = \lambda_s (I_t \otimes P_s) + \lambda_t (P_t \otimes I_s)$$