brede_mat_cov

(export/brede/brede_mat_cov.m)

Function Synopsis

C = brede_mat_cov(X, Y, varargin)

Help text

 brede_mat_cov        - Covariance matrix

       C = brede_mat_cov(X, Y)

       Input:   X    'Mat' structure or matrix
                Y    'Mat' structure or matrix

       Propery  Normalize  [ {n-1} | {n} ] Normalization 

       Output:  C    'Mat' struture with covariance

       Covariance between two matrices:

                  C = Xc'*Yc / (N-1)

       where Xc and Yc are the centered matrices of X and Y.

       Missing values (indicated with NaN) are ignored by default
       which might result in a negative definite matrix.

       Example:
         % The covariance of gaussian distributed variables
         n=1000; m=100;
         C = brede_mat_cov(randn(n,m), randn(n,m))
         figure, step=0.1; x = -4:step:4; 
         hist(C.matrix(:)*sqrt(n), x)
         line(x, brede_pdf_gauss(x)*m^2*step)

         % With missing values
         A = randn(n,m);
         A(1:n/2,1:m/2) = NaN;
         C = brede_mat_cov(A, randn(n,m))
         figure
         subplot(2,1,1)
         c = C.matrix(1:m/2,:); hist(c(:)*sqrt(n/2), x)
         line(x, brede_pdf_gauss(x)*(m/2)*m*step)
         subplot(2,1,2)
         c = C.matrix(m/2+1:end,:); hist(c(:)*sqrt(n), x)
         line(x, brede_pdf_gauss(x)*(m/2)*m*step)

       See also BREDE, BREDE_MAT, BREDE_MAT_CORRCOEF,
                BREDE_MAT_PARTCORRCOEF, BREDE_MAT_MEAN.

 $Id: brede_mat_cov.m,v 1.6 2008/03/12 14:17:01 fn Exp $

Cross-Reference Information

brede_mat_centerexport/brede/brede_mat_center.m
brede_mat_mat2matexport/brede/brede_mat_mat2mat.m