gseries (version 3.0.2)

gs.gInv_MP: Moore-Penrose inverse

Description

This function calculates the Moore-Penrose (pseudo) inverse of a square or rectangular matrix using Singular Value Decomposition (SVD). It is used internally by tsraking() and benchmarking().

Usage

gs.gInv_MP(X, tol = NA)

Value

The Moore-Penrose (pseudo) inverse of matrix X.

Arguments

X

(mandatory)

Matrix to invert.

tol

(optional)

Real number that specifies the tolerance for identifying zero singular values. When tol = NA (default), the tolerance is calculated as the product of the size (dimension) of the matrix, the norm of the matrix (largest singular value) and the machine epsilon (.Machine$double.eps).

Default value is tol = NA.

Details

The default tolerance (argument tol = NA) is coherent with the tolerance used by the MATLAB and GNU Octave software in their general inverse functions. In our testing, this default tolerance also produced solutions (results) comparable to G-Series 2.0 in SAS\(^\circledR\).

See Also

tsraking() benchmarking()

Examples

# Invertible matrix
X1 <- matrix(c(3, 2, 8, 
               6, 3, 2,
               5, 2, 4), nrow = 3, byrow = TRUE)
Y1 <- gs.gInv_MP(X1)
all.equal(Y1, solve(X1))
X1 %*% Y1

# Rectangular matrix
X2 <- X1[-1, ]
try(solve(X2))
X2 %*% gs.gInv_MP(X2)

# Non-invertible square matrix
X3 <- matrix(c(3, 0, 0, 
               0, 0, 0, 
               0, 0, 4), nrow = 3, byrow = TRUE)
try(solve(X3))
X3 %*% gs.gInv_MP(X3)