I have a matrix in which the row_names and col_names are identical. The values filling the matrix are not symmetrical (i.e. the upper triangle and the lower triangle are different). The diagonal is populated with 0s.

I want to sum the upper triangle to the lower triangle.

I don't mind what format the output is in: either a data-table of the rownames-colnames- newvalues, OR a triangular matrix (rather than a square).

The attached image shows how the col-names and row-names are equal. Eventually I want a matrix that shows e.g. for M10/M10.09 = 36 36 = 72

CodePudding user response：

Use t() to transpose the matrix, and then add to the original matrix.

Other helpful matrix functions are upper.tri(), lower.tri(), and diag(). For example, filling the upper triangle with zeros.

m<-matrix(c(0,36,3,36,0,4,1,2,0), ncol=3)

m
     [,1] [,2] [,3]
[1,]    0   36    1
[2,]   36    0    2
[3,]    3    4    0

m2<-m t(m)

m2[upper.tri(m2)]<-0

m2
     [,1] [,2] [,3]
[1,]    0    0    0
[2,]   72    0    0
[3,]    4    6    0

CodePudding user response：

M.Viking's answer has all you need if you want a matrix output, this is how you would get

a data-table of the rownames-colnames- newvalues

m<-matrix(c(0,36,3,36,0,4,1,2,0), ncol=3)
msum <- m   t(m)
msum[upper.tri(msum, diag = TRUE)] <- NA                # set diag and upper tri to NA
mdf <- as.data.frame.table(msum)                        # turn into a data frame
colnames(mdf) <- c("rownames", "colnames", "newvalues") # name df columns
mdf <- mdf[!is.na(mdf$newvalues),]                      # drop all those NA rows
mdf

  rownames colnames newvalues
2        B        A        72
3        C        A         4
6        C        B         6