nearPD {lmf}

Find nearest positive definite matrix

Description

Compute the nearest positive definite matrix to an approximate one, typically a correlation or variance-covariance matrix.

Usage

nearPD(x, corr = FALSE, keepDiag = FALSE, do2eigen = TRUE, doSym = FALSE,
doDykstra = TRUE, only.values = FALSE, only.matrix = TRUE, eig.tol = 1e-06,
conv.tol = 1e-07, posd.tol = 1e-08, maxit = 100, trace = FALSE)

Arguments

x

numeric n * n approximately positive definite matrix, typically an approximation to a correlation or covariance matrix.

corr

logical indicating if the matrix should be a correlation matrix.

keepDiag

logical, generalizing corr: if TRUE, the resulting matrix should have the same diagonal (diag(x)) as the input matrix.

do2eigen

logical indicating if a posdefify eigen step should be applied to the result of the Higham algorithm.

doSym

logical indicating if X <- (X + t(X))/2 should be done, after X <- tcrossprod(Qd, Q). Some doubt if this is necessary.

doDykstra

logical indicating if Dykstra's correction should be used; true by default. If false, the algorithm is basically the direct fixpoint iteration Y(k) = P_U(P_S(Y(k-1))).

only.values

logical; if TRUE, the result is just the vector of eigen values of the approximating matrix.

only.matrix

logical indicating if only the matrix should be returned.

eig.tol

defines relative positiveness of eigenvalues compared to largest one, \lambda_1. Eigen values \lambda_k are treated as if zero when \lambda_k / \lambda_1 = eig.tol.

conv.tol

convergence tolerance for Higham algorithm.

posd.tol

tolerance for enforcing positive definiteness (in the final posdefify step when do2eigen is TRUE).

maxit

maximum number of iterations allowed.

trace

logical or integer specifying if convergence monitoring should be traced.

Details

This function is identical to nearPD in package Matrix as far as the algorithmic method is concerned, but has an addition of the argument only.matrix to ease its application within the function fs, has lost the argument ensureSymmetry and have a small change in the list returned when only.matrix = FALSE.

Please see nearPD in package Matrix for further details.

Value

nearPD returns a numeric vector of eigen values of the approximating matrix if only.values = TRUE, returns the computed positive definite matrix if only.matrix = TRUE and else returns a list with the following componets:

mat

matrix of class "dpoMatrix", the computed positive-definite matrix.

eigenvalues

numeric vector of eigenvalues of mat.

corr

logical, just the argument corr.

normF

the Frobenius norm (norm(x-X, "F")) of the difference between the original and the resulting matrix.

iterations

number of iterations needed.

converged

logical indicating if iterations converged.

Author(s)

Jens Oehlschlaegel donated a first version. Subsequent changes by the Matrix package authors and present modifications by Thomas Kvalnes.

References

Cheng, S.H. and Higham, N. 1998. A Modified Cholesky Algorithm Based on a Symmetric Indefinite Factorization. SIAM Journal on Matrix Analysis and Applications, 19, 1097-1110.

Knol, D.L. and ten Berge, J.M.F. 1989. Least-squares approximation of an improper correlation matrix by a proper one. Psychometrika, 54, 53-61.

Higham, N. 2002. Computing the nearest correlation matrix - a problem from finance. IMA Journal of Numerical Analysis, 22, 329-343.

See Also

fs, lmf, nearPD, posdefify

Examples

#Simulated non-positive definite (PD) matrix
nonPD <- matrix(c(2.04e-03, 3.54e-05, 7.52e-03, 3.54e-05, 6.15e-07,
  1.30e-04, 7.52e-03, 1.30e-04, 2.76e-02), ncol = 3)
#View eigenvalues (PD = only positive eigenvalues)
eigen(nonPD)
#Calculate PD matrix
PD <- nearPD(nonPD, only.matrix = TRUE)
PD
#View eigenvalues
eigen(PD)
#More thorough examples are given in the help pages for nearPD
 #in the Matrix package.