R：用不同大小的矩阵列表生成矩阵

Question

I have a list of 1,000 matrices, whose two first elements are:

> Bpam[1:2]
[[1]]
       cluster sil_width
D.var        1 0.7445190
Hy.adu       1 0.7233527
A.cra        0 0.8563551
Cu.cir       0 0.8483707
Sp.sp        0 0.8461553
E.gad        0 0.8368920
L.elo        0 0.8341050
A.mor        0 0.8219688
H.com        0 0.7046171
S.cad        0 0.5731629

[[2]]
       cluster sil_width
Hy.adu       1 0.5518385
D.var        1 0.1878755
Ab.gad       0 0.8183177
L.elo        0 0.7964595
A.cra        0 0.7964595
Cu.cir       0 0.7879583
A.mor        0 0.7667134
S.cad        0 0.5821773
H.com        0 0.5644358

Note that the number of rows differ ("Sp.sp" is absent in Bpam[[2]]. I would like to create a matrix of n columns and 1,000 rows (where n is the total number of different rownames across the list) to save the variable cluster of each matrix in the list and NAs if one of the rows is missing. I tried

NA.matrix <- matrix(rep(NA, n*length(Bpam)), length(Bpam), n)
colnames(NA.matrix) <- A # char vector with names in n
# 
clus.memb.p <- sapply(1:length(Bpam), function(x) 
               NA.matrix[x, which(colnames(NA.matrix) %in% rownames(Bpam[[x]]))] 
               <- Bpam[[x]][,1])  

but its does not return a matrix. Any help will be most welcome.

Answer 1

We can merge by the row.names of the list of matrices . With all = TRUE , it will create NA by default if one of the row names is missing in the matrix

out <- Reduce(function(...) merge(..., by = "row.names", all = TRUE), Bpam)
row.names(out) <- out$Row.names
out <- out[, -1]

data

Bpam <- list(structure(c(1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0.744519, 0.7233527, 
0.8563551, 0.8483707, 0.8461553, 0.836892, 0.834105, 0.8219688, 
0.7046171, 0.5731629), .Dim = c(10L, 2L), .Dimnames = list(c("D.var", 
"Hy.adu", "A.cra", "Cu.cir", "Sp.sp", "E.gad", "L.elo", "A.mor", 
"H.com", "S.cad"), c("cluster", "sil_width"))), structure(c(1, 
1, 0, 0, 0, 0, 0, 0, 0.5518385, 0.1878755, 0.8183177, 0.7964595, 
0.7964595, 0.7879583, 0.7667134, 0.5821773), .Dim = c(8L, 2L), .Dimnames = list(
    c("Hy.adu", "D.var", "Ab.gad", "L.elo", "A.cra", "Cu.cir", 
    "A.mor", "S.cad"), c("cluster", "sil_width"))))

Answer 2

I found a very simple solution. First, I modified my function to generate a list of matrices ordered alphabetically by row.names:

> Bpam[1:2]
[[1]]
       cluster  sil_width
A.cra        0 0.81960918
A.mor        0 0.83767035
Cu.cir       0 0.82537241
D.var        0 0.05898329
H.com        0 0.55435303
Hy.adu       1 0.00000000
L.elo        0 0.83984674
S.cad        0 0.81047726
Sp.sp        0 0.80335860

[[2]]
       cluster  sil_width
A.cra        0 0.85243545
A.mor        0 0.88501435
Cu.cir       0 0.87372261
D.var        0 0.02317251
E.gad        0 0.85797643
H.com        0 0.89236411
Hy.adu       1 0.00000000
S.cad        0 0.88181818
Sp.sp        0 0.88836713

Then

NA.matrix <- matrix(rep(NA, n*length(Bpam)), length(Bpam), n)
colnames(NA.matrix) <- sort(A)

So both the NA matrix and elements in the list are arranged alphabetically by col and row names, respectively. Now I use a loop to fill in the NA matrix:

for (i in 1:length(Bpam)) {
  NA.matrix[i, which(colnames(NA.matrix) %in% rownames(Bpam[[i]]))] <- Bpam[[i]][,1] 
  }

Done!