ltsa (version 1.4.6)

TrenchInverse: compute the matrix inverse of a positive-definite Toepliz matrix

Description

The Trench algorithm (Golub and Vanload, 1983) is implemented in C and interfaced to R. This provides an expedient method for obtaining the matrix inverse of the covariance matrix of n successive observations from a stationary time series. Some applications of this are discussed by McLeod and Krougly (2005).

Usage

TrenchInverse(G)

Arguments

G

a positive definite Toeplitz matrix

Value

Warning

You should test the input x using is.toeplitz(x) if you are not sure if x is a symmetric Toeplitz matix.

References

Golub, G. and Van Loan (1983). Matrix Computations, 2nd Ed. John Hoptkins University Press, Baltimore. Algorithm 5.7-3.

McLeod, A.I., Yu, Hao, Krougly, Zinovi L. (2007). Algorithms for Linear Time Series Analysis, Journal of Statistical Software.

See Also

TrenchLoglikelihood, is.toeplitz, DLLoglikelihood, TrenchMean, solve

Examples

#compute inverse of matrix and compare with result from solve
data(LakeHuron)
r<-acf(LakeHuron, plot=FALSE, lag.max=4)$acf
R<-toeplitz(c(r))
Ri<-TrenchInverse(R)
Ri2<-solve(R)
Ri
Ri2

#invert a matrix of order n and compute the maximum absolute error
# in the product of this inverse with the original matrix
n<-5	   
r<-0.8^(0:(n-1))
G<-toeplitz(r)
Gi<-TrenchInverse(G)
GGi<-crossprod(t(G),Gi)
id<-matrix(0, nrow=n, ncol=n)
diag(id)<-1
err<-max(abs(id-GGi))
err