Block diagonal matrices in R

April 13, 2011

By CL

(This article was first published on Christophe Ladroue » R, and kindly contributed to R-bloggers)

As far as I can tell, R doesn’t have a function for building block diagonal matrices so as I needed one, I’ve coded it myself. It might save someone some time.

Example:
Let m1 and m2 two square matrices.

Selec All Code:

1 2	m1=matrix(runif(1010),nrow=10,ncol=10) m2=matrix(runif(55),nrow=5,ncol=5)

By passing any number of matrices as argument:

Selec All Code:

1	blockMatrix<-blockMatrixDiagonal(m1,m2,m2,m1)

Or a list of matrices:

Selec All Code:

1	blockMatrix<-blockMatrixDiagonal(list(m1,m2,m2,m1))

It produces:

Selec All Code:

1 2	colourScale<-seq(0,1,length.out=100) image(blockMatrix[30:1,],asp=1,col=rgb(colourScale,colourScale,colourScale),ann=FALSE,xaxt="n",yaxt="n",axes=FALSE)

That is, a matrix which is zero everywhere but with m1,m2,m2 and m1 on the diagonal.
Code:

Selec All Code:

# builds a block matrix whose diagonals are the square matrices provided.
# m1=matrix(runif(10*10),nrow=10,ncol=10)
# m2=matrix(runif(5*5),nrow=5,ncol=5)
# blockMatrix<-blockMatrixDiagonal(m1,m2,m2,m1)
# or
# blockMatrix<-blockMatrixDiagonal(list(m1,m2,m2,m1))
# C.Ladroue
 
blockMatrixDiagonal<-function(...){  
  matrixList<-list(...)
  if(is.list(matrixList[[1]])) matrixList<-matrixList[[1]]
 
  dimensions<-sapply(matrixList,FUN=function(x) dim(x)[1])
  finalDimension<-sum(dimensions)
  finalMatrix<-matrix(0,nrow=finalDimension,ncol=finalDimension)
  index<-1
  for(k in 1:length(dimensions)){
    finalMatrix[index:(index+dimensions[k]-1),index:(index+dimensions[k]-1)]<-matrixList[[k]]
    index<-index+dimensions[k]
    }
    finalMatrix
  }