org.akutan.correlations
Class SpectralDecomposition

java.lang.Object
  org.akutan.correlations.SpectralDecomposition

public class SpectralDecomposition
extends java.lang.Object

Implements the Spectral Decomposition method from the paper "The Most General Way to Create a Valid Correlation Matrix for Risk Management and Option Pricing Purposes" by Rebonato and Jaekel, 1999. It is available at http://www.quarchome.org/correlationmatrix.pdf

Since:

18 Dec 2008

Method Summary

static cern.colt.matrix.DoubleMatrix2D

adjust(cern.colt.matrix.DoubleMatrix2D input) This method is from the paper "The Most General Way to Create a Valid Correlation Matrix for Risk Management and Option Pricing Purposes" by Rebonato and Jaekel, 1999.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

adjust

public static cern.colt.matrix.DoubleMatrix2D adjust(cern.colt.matrix.DoubleMatrix2D input)

This method is from the paper "The Most General Way to Create a Valid Correlation Matrix for Risk Management and Option Pricing Purposes" by Rebonato and Jaekel, 1999.

The algorithm is roughly as follows:

• Compute the eigenvalues
• Set all negative eigenvalues to 0
• Compute the scaling factor
• Reconstruct the matrix using the scaling factor

Parameters:

input - Correlation matrix to be cleaned up

Returns:

Transformed correlation matrix which is now positive semi-definite and a valid correlation matrix (diagonal 1's and all elements on the interval (-1, 1).