Apply function over every entry one table to every entry of another

Question

I would like to apply a function, bandedlossfn to all entries in loss.tib using every item in bandstib

library(tidyverse)
set.seed(1)
n <- 5
loss <- rbeta(n, 1, 10) * 100

loss.tib <- loss %>% as_tibble %>% mutate(loss = value) %>% mutate(lossid = 
row_number()) %>% select(lossid, loss)
bandstib <- tibble(bandid = seq(4),
                   start = seq(0, 75, by = 25),
                   end = seq(25, 100, by = 25))

bandedlossfn <- function(loss, start, end) {
  pmin(end - start, pmax(0, loss - start))
}

As per the answer below the following code produces the calculation:

loss.tib %>% 
mutate(
  result = map(
    loss, ~ tibble(result = bandedlossfn(.x, bandstib$start, bandstib$end))
    )
    ) %>% unnest

However I'd like to include the bandid as the index within the map function and additionally filter(!near(result,0)) with the map function.

The result should be:

lossid  loss    bandid  result
1   21.6691088  1   21.6691088  
2   6.9390647   1   6.9390647   
3   0.5822383   1   0.5822383   
4   5.5671643   1   5.5671643   
5   27.8237244  1   25.0000000  
5   27.8237244  2   2.8237244

Sorry, should be loss – Vincent Risington Apr 21 '19 at 14:57 — Vincent Risington, Apr 21 '19 at 14:57
and what would your expected output look like? – Ronak Shah Apr 21 '19 at 15:00 — Ronak Shah, Apr 21 '19 at 15:00
@ Ronak Shah - added – Vincent Risington Apr 21 '19 at 15:24 — Vincent Risington, Apr 21 '19 at 15:24

Cettt · Accepted Answer · 2019-04-21T14:42:56.767

1

here is one possibility using map2 from the purrr package:

bandstib %>% 
  mutate(result = map2(start, end, ~bandedlossfn(loss.tib[[1]], .x, .y)))

Depending on how you want your output to be you can continue from there by e.g. using unnest.

EDIT

Here is how you can apply it the otherway arround using map instead of map2:

loss.tib %>% 
  mutate(result = map(value, bandedlossfn, start = bandstib$start, end = bandstib$end)) %>%
  unnest() %>% 
  mutate(bandid = rep(seq(4), n))

edited Apr 21 '19 at 14:42

answered Apr 21 '19 at 14:20

Cettt

11,460
7
35
58

Hello @Cett, thanks how can I get this the other way round? I want to apply the bands to the loss table – Vincent Risington Apr 21 '19 at 14:23
hi, sorry, i misunderstood your question. I updated my answer – Cettt Apr 21 '19 at 14:26
that's great! how can I include the `bandid` when I unnest? – Vincent Risington Apr 21 '19 at 14:39
I included this to. Depending on your exact data it would probably make more sense to include the bandid as an output `bandedlossfn`. – Cettt Apr 21 '19 at 14:43
Ok this does it. However, my loss.tib has 30m rows so I'd like to index and filter results != 0 within the map function itself. Not sure if you have a quick solution or if I should ask as another question. – Vincent Risington Apr 21 '19 at 15:26
probably ask another question or modify this question. – Cettt Apr 21 '19 at 15:28
Let us [continue this discussion in chat](https://chat.stackoverflow.com/rooms/192171/discussion-between-cettt-and-vincent-risington). – Cettt Apr 21 '19 at 15:30

Apply function over every entry one table to every entry of another

1 Answers1

Linked