fdlm_init: Initialize the reduced-rank functional data model

In drkowal/dfosr: Dynamic Function-on-Scalars Regression

Description

Compute initial values for the factors and loadings curves, where the input matrix may have NAs.

Usage

fdlm_init(Y, tau, K = NULL, use_pace = FALSE)

Arguments

Y	the T x m data observation matrix, where T is the number of time points and m is the number of observation points (NAs allowed)
tau	the m x d matrix of coordinates of observation points
K	the number of factors; if NULL, select using proportion of variability explained (0.99)
use_pace	logical; if TRUE, use the PACE procedure for FPCA (only for d = 1); otherwise use splines

Details

The PACE procedure is useful for estimated a reduced-rank functional data model in the case of sparsely-observed FD, i.e., many NAs in Y. However, it is also only valid for univariate observation points

Value

a list containing

• Beta the T x K matrix of factors

• Psi the J x K matrix of loading curve basis coefficients, where J is the number of spline basis functions

• splineInfo a list with information about the spline basis

• Y0 the imputed data matrix

drkowal/dfosr documentation built on May 7, 2020, 3:09 p.m.