Replace NaN's of one matrix with values of another matrix in R

Question

I have a 100 x 100 matrix that has 1800 missing values represented by NaN . I have another 1800 x 1 matrix that contains all these missing values. I want to plug in these values from my 1800 x 1 matrix to the first matrix.

I tried;

data_matrix[is.na(data_matrix)] <- predicted_values[1,]

But this did not work.

Perhaps the safest option would be to do it based on row/col names? Because both of my matrices have corresponding row/col names. Here's an example of what I mean:

data_matrix

  VAL.T  VAL.U  VAL.V  VAL.W
A 10.5   NaN    203    902
B 20.9   343    12     NaN
C 32     22     NaN    90
D 12     NaN    NaN    23.1

predicted_values

VAL.U:A 65
VAL.W:B 21
VAL.V:C 23.9
VAL.U:D 11.1
VAL.V:D 78

The order that values are arranged in predicted_values might also be helpful: ie values in predicted_values are arranged in a way as if we're moving from the beginning of first row in data_matrix to the end of first_row and then moving to next row. This is the way NaN values should be replaced.

Answer 1

Since predicted_values stores the values by row then column, while matrices normally store their values column-wise, you should do transposition.

a=t(data_matrix)
a[is.na(a)]=predicted_values[,1]
data_matrix=t(a)

Answer 2

Use a two-column matrix built from splitting the first predval column as the first (single) argument to "[<-" and predval's second column as the second argument:

idxm <- cbind( sapply( strsplit(as.character(predval$V1), ":"), "[",2), 
                 sapply( strsplit(as.character(predval$V1), ":"), "[",1) )
dtm[ idxm ] <- predval$V2
dtm
#-----------
  VAL.T VAL.U VAL.V VAL.W
A  10.5  65.0 203.0 902.0
B  20.9 343.0  12.0  21.0
C  32.0  22.0  23.9  90.0
D  12.0  11.1  78.0  23.1



dput(predval)
structure(list(V1 = structure(c(1L, 5L, 3L, 2L, 4L), .Label = c("VAL.U:A", 
"VAL.U:D", "VAL.V:C", "VAL.V:D", "VAL.W:B"), class = "factor"), 
    V2 = c(65, 21, 23.9, 11.1, 78)), .Names = c("V1", "V2"), class = "data.frame", row.names = c(NA, 
-5L))

dput(dtm)
structure(c(10.5, 20.9, 32, 12, 65, 343, 22, 11.1, 203, 12, 23.9, 
78, 902, 21, 90, 23.1), .Dim = c(4L, 4L), .Dimnames = list(c("A", 
"B", "C", "D"), c("VAL.T", "VAL.U", "VAL.V", "VAL.W")))