CODEWITHSUNDEEP
rloopsdplyr
MCS
MCS

Reputation: 1101

Loop within mutate

I calculate an index using dplyr. The index is the summation of the squared ratios between each entry and the total entry in a group.

library(dplyr)

set.seed(1e2)
firm_id <-  sample(1:3, 1e2, rep=T)
pro_id <-  sample(1:8, 1e2, rep=T)
emplo_id <- sample(1:5, 1e2, rep=T)
cost <-  round(abs(rnorm(1e2, 20)), 2)

df <- data.frame(firm_id, pro_id, emplo_id, cost)

df_index <- df %>% group_by(firm_id,pro_id) %>% 
  mutate(INDEX = sum((cost/sum(cost))^2))

I want now to calculate how much each entry contributes to the idex its group produces, meaning that I want to calculate a new index as if the entry cost for a value were 0, and this for every entry as if in a loop (then divide the new index by the old).

Expected results:

firm_id <-  c(1,1,1)
pro_id <-  c(1,1,1)
emplo_id <- c(1:3)
cost <-  c(1,50,100)
INDEX <- rep(0.5482654,3)
newINDEX <- c(0.5555556,0.9803941,0.9615532)
df_index <- data.frame(firm_id, pro_id, emplo_id, cost, INDEX, newINDEX)

With mutate I have no idea how to do it. Any suggestion welcome!

Upvotes: 2

Views: 1291

Answers (1)

Mikko Marttila
Mikko Marttila

Reputation: 11878

You can use purrr::map_dbl() to loop over the row indices within each group, and then apply a function that replaces the cost at a given index with 0 and then recalculates the index. Here's an example with the data that you gave the expected output for:

library(dplyr)
library(purrr)

# The function used to calculate the index value
index <- function(x) sum((x / sum(x)) ^ 2)

df_index %>%
  group_by(firm_id, pro_id) %>%
  mutate(new = map_dbl(row_number(), function(i) {
    index(replace(cost, i, 0))
  }))
#> # A tibble: 3 x 7
#> # Groups:   firm_id, pro_id [1]
#>   firm_id pro_id emplo_id  cost INDEX newINDEX   new
#>     <dbl>  <dbl>    <int> <dbl> <dbl>    <dbl> <dbl>
#> 1       1      1        1     1 0.548    0.556 0.556
#> 2       1      1        2    50 0.548    0.980 0.980
#> 3       1      1        3   100 0.548    0.962 0.962

With an additional helper function, you can also do this in a cleaner fashion: 

index_without <- function(i, x) {
  map_dbl(i, function(i) index(replace(x, i, 0)))
}

df_index %>%
  group_by(firm_id, pro_id) %>%
  mutate(new = index_without(row_number(), cost))
#> # A tibble: 3 x 7
#> # Groups:   firm_id, pro_id [1]
#>   firm_id pro_id emplo_id  cost INDEX newINDEX   new
#>     <dbl>  <dbl>    <int> <dbl> <dbl>    <dbl> <dbl>
#> 1       1      1        1     1 0.548    0.556 0.556
#> 2       1      1        2    50 0.548    0.980 0.980
#> 3       1      1        3   100 0.548    0.962 0.962

Created on 2018-08-08 by the reprex package (v0.2.0.9000).

Upvotes: 2

Related Questions