sparse_lm: Fit Linear Model using sparse matrices

In tstev/tstevR: Someone's Collection of R functions

Description

sparse_lm uses sparse matrices and a iterative solver to fit linear models. The sparse iterative solvers are from the Eigen C++ library. Currently, only the Conjugate Gradient algorithm has been implemented with a diagonal preconditioner.

Usage

sparse_lm(X, y, x0, maxiter = 4L * ncol(X), tol = .Machine$double.eps)

Arguments

X	a sparse matrix of the class dgCMatrix-class.
y	a numeric vector with response variable.
x0	a numeric vector of length ncol(X) with an initial guess. If it's not provided it will default to a zero-vector.
maxiter	an integer scalar with the maximum number of iterations. A good choice is usually 4 * ncol(X).
tol	a numeric scalar with the relative error tolerance.

Value

A list with the elements

coefficients	The estimated coefficients from solver.
itr	A scalar denoting the number of iterations.
error	A scalar denoting the the relative error at the last iteration.
fitted.values	A vector with the fitted values.
residuals	The vector with the residuals.

See Also

See Eigen C++ library for documentation on Eigen library and Conjugate Gradient for more information about this method.

tstev/tstevR documentation built on May 18, 2020, 10:57 p.m.