MSVD is the jSpectral module of jLab.

 MSVD  Singular value decomposition for polarization analysis.
 
    MSVD computes the singular value decomposition for multiple-
    transform polarization analysis at all or selected frequencies.
 
    [D,U1,V1]=MSVD(W) computes the singular value decomposition of 
    sub-matrices of the eigentransform W.
 
    The input matrix W is an J x N x K matrix where 
 
           J is the number of transforms (frequencies or scales)
           N is the number of dataset components
           K is the number of eigentransforms (tapers or wavelets)
 
    The output is
 
 	       D  --  J x MIN(K,N) matrix of singular values
 	      U1  --  J x N matrix of first left singular vectors
 	      V1  --  J x K matrix of first right singular vectors
 
    The left singular vectors U1 are the eigenvectors of the spectral
    matrix, while the right singular vectors V1 are the eigenvectors of 
    the "structure matrix".  
 
    W may optionally be of size M x J x N x K, in which case the
    output arguments are all three-dimensional matrices with sizes
 
 	       D  --  M x J x MIN(K,N) matrix of singular values
 	      U1  --  M x J x N matrix of first left singular vectors
 	      V1  --  M x J x K matrix of first right singular vectors
 
    [D,U1,V1]=MSVD(W,INDEX) for three-dimensional W optionally 
    performs the SVD only at the rows of MMAT indicated by INDEX.  The 
    output arguments then all have LENGTH(INDEX) rows rather than M.
 
    [D,U1,V1,TR]=MSVD optionally outputs TR, the trace of the spectral 
    matrix, which is of size J (for 3-D W) or M x J (for 4-D W).  
 
    MSVD(..., 'quiet') suppresses a progress report.
 	
    Usage:  [d,u1,v1]=msvd(w);
            [d,u1,v1]=msvd(w,index);
            [d,u1,v1,tr]=msvd(w,index);
 
    'msvd --t' runs a test.
    _________________________________________________________________
    This is part of JLAB --- type 'help jlab' for more information
    (C) 1993--2014 J.M. Lilly --- type 'help jlab_license' for details