How to create dummy variables that indicate the presence of a factor for other observations within in a group?

Question

I am working with a data frame like the following, where Color and `Player are factor variables:

I want to create indicator variables for each value of the color column. However, I want those indicator variables to represent whether the color is present for other players in the same game (not whether it's present for that player). So I want the above table to turn into:

I imagine the code will have group_by(Game) %>%, but I'm lost beyond that.

Data:

structure(list(Game = c("A", "A", "A", "B", "B", "B"), Player = c(1L, 
2L, 3L, 1L, 2L, 3L), Color = c("Red", "Green", "Blue", "Green", 
"Purple", "Yellow"), Blue = c(1L, 1L, 0L, 0L, 0L, 0L), Green = c(1L, 
0L, 1L, 0L, 1L, 1L), Yellow = c(0L, 0L, 0L, 1L, 1L, 0L), Red = c(0L, 
1L, 1L, 0L, 0L, 0L), Purple = c(0L, 0L, 0L, 1L, 0L, 1L)), class = "data.frame", row.names = c(NA, 
-6L))

Answer 1

Perhaps this helps - split the 'Color' column by 'Game', create a binary matrix by comparing the elements of 'Color' (!=), convert to tibble, row bind (_dfr) and bind the dataset with the original dataset (bind_cols)

library(purrr)
library(dplyr)
library(tidyr)
map_dfr(split(df1$Color, df1$Game), ~ {
   m1 <- +(outer(.x, .x, FUN = `!=`))
  colnames(m1) <- .x
  as_tibble(m1)}) %>%
  mutate(across(everything(), replace_na, 0)) %>%
  bind_cols(df1, .)

-output

  Game Player  Color Red Green Blue Purple Yellow
1    A      1    Red   0     1    1      0      0
2    A      2  Green   1     0    1      0      0
3    A      3   Blue   1     1    0      0      0
4    B      1  Green   0     0    0      1      1
5    B      2 Purple   0     1    0      0      1
6    B      3 Yellow   0     1    0      1      0

---

Or another option is with dummy_cols and then modify the output

library(fastDummies)
library(stringr)
dummy_cols(df1, 'Color') %>%
  rename_with(~ str_remove(.x, "Color_")) %>%
  group_by(Game) %>% 
  mutate(across(Blue:Yellow, ~ +(Color != cur_column() & any(.x)))) %>%
   ungroup

-output

# A tibble: 6 × 8
  Game  Player Color   Blue Green Purple   Red Yellow
  <chr>  <int> <chr>  <int> <int>  <int> <int>  <int>
1 A          1 Red        1     1      0     0      0
2 A          2 Green      1     0      0     1      0
3 A          3 Blue       0     1      0     1      0
4 B          1 Green      0     0      1     0      1
5 B          2 Purple     0     1      0     0      1
6 B          3 Yellow     0     1      1     0      0

data

df1 <- structure(list(Game = c("A", "A", "A", "B", "B", "B"), Player = c(1L, 
2L, 3L, 1L, 2L, 3L), Color = c("Red", "Green", "Blue", "Green", 
"Purple", "Yellow")), row.names = c(NA, -6L), class = "data.frame")

Answer 2

Here is a way how we could do it:

First we use model.matrix() fucntion multiply it by 1 and substract 1 within a wrap of abs(). Then we get almost the desired output, the only thing that is left is the get zeros in case if non of the colors is present. We do this with a mutate across...:

library(dplyr)
df %>% 
  cbind(abs((model.matrix(~ Color + 0, .) == 1)*1-1)) %>% 
  group_by(Game) %>% 
  mutate(across(-c(Player, Color), ~case_when(sum(.)==3 ~0, 
                                              TRUE ~ .)))

  Game  Player Color  ColorBlue ColorGreen ColorPurple ColorRed ColorYellow
  <chr>  <int> <chr>      <dbl>      <dbl>       <dbl>    <dbl>       <dbl>
1 A          1 Red            1          1           0        0           0
2 A          2 Green          1          0           0        1           0
3 A          3 Blue           0          1           0        1           0
4 B          1 Green          0          0           1        0           1
5 B          2 Purple         0          1           0        0           1
6 B          3 Yellow         0          1           1        0           0
>