I have the following table:

name	group
a	1
b	1
c	2
d	2
e	3
f	3

and I want to randomly re-assign the group membership by (i) making sure that the names will not be assigned to same group, and (ii) the probability of the group membership will remain the same. Also, I am trying to (iii) avoid all the names under the same group to be assigned to the same new group. In essence I want to achieve something like this:

name	group	new.group
a	1	2
b	1	3
c	2	1
d	2	3
e	3	1
f	3	2

How do I do this in R?

CodePudding user response：

A base R option using sample setdiff

transform(
  df,
  new.group = ave(group, group, FUN = function(x) sample(setdiff(group, x),length(x)))
)

gives

  name group new.group
1    a     1         2
2    b     1         3
3    c     2         1
4    d     2         3
5    e     3         1
6    f     3         2

Data

> dput(df)
structure(list(name = c("a", "b", "c", "d", "e", "f"), group = c(1L, 
1L, 2L, 2L, 3L, 3L)), class = "data.frame", row.names = c(NA,
-6L))

CodePudding user response：

With all the restrictions, this is barely shuffling anymore: You could use the modulo operator.

df %>%
  group_by(group) %>%
  mutate(new_group = (2   row_number()   group) %% 3   1)

  name  group new_group
  <chr> <int>     <dbl>
1 a         1         2
2 b         1         3
3 c         2         3
4 d         2         1
5 e         3         1
6 f         3         2