2D Function-on-Scalar Regression — fosr.2d • fdars

Fits a penalized regression model where the response is a 2D surface Y_i(s,t) and predictors are scalar.

Usage

fosr.2d(fdataobj, predictors, argvals.s, argvals.t, lambda.s = 0, lambda.t = 0)

Arguments

fdataobj: An fdata object where each row is a flattened 2D surface (m1 * m2 grid points).
predictors: A matrix of scalar predictors (n x p).
argvals.s: Grid points along the s dimension.
argvals.t: Grid points along the t dimension.
lambda.s: Smoothing penalty in s direction (default 0).
lambda.t: Smoothing penalty in t direction (default 0).

Value

An object of class 'fosr.2d' with components:

intercept: Intercept surface as fdata
beta: Coefficient surfaces as fdata
fitted: Fitted surfaces as fdata
residuals: Residual surfaces as fdata
r.squared: Global R-squared
r.squared.pointwise: Pointwise R-squared values
lambda.s: Penalty in s direction
lambda.t: Penalty in t direction
gcv: GCV criterion
grid: List with argvals.s, argvals.t, m1, m2

Examples

# \donttest{
n <- 30; m1 <- 5; m2 <- 5
fd <- fdata(matrix(rnorm(n * m1 * m2), n, m1 * m2))
X <- matrix(rnorm(n * 2), n, 2)
fit <- fosr.2d(fd, X, seq(0, 1, length.out = m1), seq(0, 1, length.out = m2))
# }