vech {lazy.mat} | R Documentation |
Vectorization of the lower-half or upper-half of a square matrix
vech(S, nodiag = 0, byrow = 1, lowerh = 1)
S |
Input square matrix or array whose lower or upper half elements will be vectorized. |
nodiag |
= 1 to omit the diagonal elements |
byrow |
= 0 to vectorize in col-major order |
lowerh |
= 0 to vectorize upper-half |
Let M be 4 x 4.
When byrow=1 and lowerh=1,
vech(M)=c( m11, m21, m22, m31, m32, m33, ..., mn1, mn2, ..., mnn )
When byrow=1 and lowerh=0,
vech(M)=c( m11, m12, m13, ..., m1n, m22, m23, m24, ..., m2n, ..., mnn )
When byrow=0 and lowerh=1, the upper half of the matrix will be vectorized in row major order as:
vech(M)=c( m11, m21, m31, ..., mn1, m22, m32, ..., mn2, m33, ...., mnn )
When byrow=0 and lowerh=0, the upper half of the matrix will be vectorized in row major order as:
vech(M)=c( m11, m12, m22, m13, m23, m33, m14, ..., m44, m1n, m2n, ...., mnn )
Note that vech(M,lowerh=0) and vech(t(M),lowerh=1) are the same
except for their row names.
Use vechinv
to recover the original matrix.
A column vector when S is a matrix, or a matrix when S is an array.
ndim=5 M=demomat(ndim,ndim,root="m") cat("nodiag=0: byrow=1, lower=1 and byrow=0, lower=1\n") Print(vech(M,0,1,1),vech(M,0,0,1)) cat("nodiag=0: byrow=1, lower=0 and byrow=0, lower=0\n") Print(vech(M,0,1,0),vech(M,0,0,0)) cat("nodiag=1: byrow=1, lower=1 and byrow=0, lower=1\n") Print(vech(M,1,1,1),vech(M,1,0,1)) cat("nodiag=1: byrow=1, lower=0 and byrow=0, lower=0\n") Print(vech(M,1,1,0),vech(M,1,0,0)) cat("Should be equal.") Print(vech(M,lowerh=0), vech(t(M),byrow=0)) # converting to a native vector with name vecM=vech(M) vecM=vecM[,1,drop=1]