Is there a possibility to merge several origin destination matrices into one dataframe?

83 views Asked by At

I have created origin destination matrices for different weeks in the year, e.g. the output looks like:

Region 1 Region 2 Region 3
Region 1 0 8 1
Region 2 4 3 3
Region 3 2 2 3

Week 1

I have similar looking matrices for all weeks of the year, all representing activity between each pair of nodes. Now, I want to compute a dataframe which shows activity for all different pairs of origin-destination (13x13) per week in the year. How can I code this using R?

1

There are 1 answers

1
Allan Cameron On BEST ANSWER

Obviously we don't have your data. I'll create a little example data set here so you can see one approach that should work for you.

Suppose I have three matrices representing three weeks:

mat1
#>          Region 1 Region 2 Region 3
#> Region 1        0        8        1
#> Region 2        4        3        3
#> Region 3        2        2        3

mat2
#>          Region 1 Region 2 Region 3
#> Region 1        9        6        2
#> Region 2        3        4        7
#> Region 3        5        8        1

mat3
#>          Region 1 Region 2 Region 3
#> Region 1        6        8        5
#> Region 2        9        3        1
#> Region 3        7        4        2

(The code to recreate these matrices is shown at the bottom of this answer in a format you can copy and paste to your R console).

The first thing to do is to get all your matrices into a list (if they are not already)

my_list <- list(mat1, mat2, mat3)

Now you can melt the matrices into data frames. Rather than doing this one at a times, we can do them all at once now that they are in a list by calling lapply:

library(reshape2)

my_dfs  <- lapply(my_list, melt)

This will give us a list of data frames, one for each week. Now we need to bind these together into a single long data frame.

df      <- do.call(rbind, my_dfs)

Lastly, we want to add an extra column to the data frame so that we know which week the data comes from:

df$week <- rep(seq(length(my_list)), each = length(mat1))

And this gives us the final result:

df
#>        Var1     Var2 value week
#> 1  Region 1 Region 1     0    1
#> 2  Region 2 Region 1     4    1
#> 3  Region 3 Region 1     2    1
#> 4  Region 1 Region 2     8    1
#> 5  Region 2 Region 2     3    1
#> 6  Region 3 Region 2     2    1
#> 7  Region 1 Region 3     1    1
#> 8  Region 2 Region 3     3    1
#> 9  Region 3 Region 3     3    1
#> 10 Region 1 Region 1     9    2
#> 11 Region 2 Region 1     3    2
#> 12 Region 3 Region 1     5    2
#> 13 Region 1 Region 2     6    2
#> 14 Region 2 Region 2     4    2
#> 15 Region 3 Region 2     8    2
#> 16 Region 1 Region 3     2    2
#> 17 Region 2 Region 3     7    2
#> 18 Region 3 Region 3     1    2
#> 19 Region 1 Region 1     6    3
#> 20 Region 2 Region 1     9    3
#> 21 Region 3 Region 1     7    3
#> 22 Region 1 Region 2     8    3
#> 23 Region 2 Region 2     3    3
#> 24 Region 3 Region 2     4    3
#> 25 Region 1 Region 3     5    3
#> 26 Region 2 Region 3     1    3
#> 27 Region 3 Region 3     2    3

Created on 2022-03-11 by the reprex package (v2.0.1)


Data

mat1 <- structure(c(0L, 4L, 2L, 8L, 3L, 2L, 1L, 3L, 3L), .Dim = c(3L, 
3L), .Dimnames = list(c("Region 1", "Region 2", "Region 3"), 
    c("Region 1", "Region 2", "Region 3")))

mat2 <- structure(c(9L, 3L, 5L, 6L, 4L, 8L, 2L, 7L, 1L), .Dim = c(3L, 
3L), .Dimnames = list(c("Region 1", "Region 2", "Region 3"), 
    c("Region 1", "Region 2", "Region 3")))

mat3 <- structure(c(6L, 9L, 7L, 8L, 3L, 4L, 5L, 1L, 2L), .Dim = c(3L, 
3L), .Dimnames = list(c("Region 1", "Region 2", "Region 3"), 
    c("Region 1", "Region 2", "Region 3")))