Build Sparse Tensor Design Matrix — build_tensor_matrix

Inner function to construct the tensor product design matrix as a sparse SparseM::matrix.csr, exploiting the local support of the B-spline basis (each row has only a handful of non-zero entries). The column layout is identical to cbind(1, build_tensor_matrix(...)), i.e. it includes the leading global intercept column, so the result can be passed directly to quantreg::rq.fit.sfn.

Usage

build_tensor_matrix_sparse(scores, Y.mask, spline.basis)

Arguments

scores: A matrix of functional Principal Component Analysis (fPCA) scores for each subject.
Y.mask: A logical mask matrix of dimensions (N x T) indicating which time points are observed (TRUE) or missing (FALSE) for each subject.
spline.basis: The evaluation matrix of the B-spline basis functions at the time points.

Value

A matrix.csr sparse matrix containing the intercept column, the intercept bases and the scaled spline bases for all valid observations.