matrix-scalar multiplication within dataframes in R

Question

I have a dataframe containing cost data and a separate list of matrices that state how each type of cost should be allocated. For the sake of explanation, I'll make some dummy data.

library(tidyverse)

# here is my list of cost allocation matrices
matrList <- list(A = cbind(runif(3),runif(3),runif(3)), 
                 B = cbind(runif(3),runif(3),runif(3)), 
                 C = cbind(runif(3),runif(3),runif(3))) %>% 
  lapply(function(x){
    rownames(x) <- c("x","y","z")
    colnames(x) <- c("p","q","r")
    x <- x/sum(x)
  })

# and here is my cost data
mydf <- data.frame(val = runif(8,0,100),
                   lookup = c("A","C","B","B","C","B","C","A"))

So each cost is either type A, type B or type C. For each of these types, there is a 3x3 matrix that sums to 1 that states how that cost should be allocated.

I've managed to botch a kind of left_join that adds a column to the dataframe containing the appropriate matrices

mydf <- mydf %>%
    mutate(propMatrix = lapply(lookup, function(x) matrList[[x]]))

This seems to keep the structure of the matrices - great. Only now I want to multiply each propMatrix by the corresponding val in the dataframe to give me matrices containing the actual allocated costs rather than just the proportions. Optimistically, I thought something simple like mutate(valMatrix = val*propMatrix) might work but no.

Does anyone have any smart ideas? I suppose it could be done in a loop as a last resort, but it's not hugely practical given the size of my actual data. I also don't think it's practical to abandon approach of having matrices within the dataframe and put it in long form with replicated rows since each matrix is ~600x100.

Any thoughts much appreciated.

Thanks in advance, James

akrun akrun · Accepted Answer · 2017-11-22T15:33:53

We may need map2

mydf %>%
    mutate(valMatrix = map2(propMatrix, val, `*`))

matrix-scalar multiplication within dataframes in R

1 Answers