dplyr 唯一行 sample_n

Question

I am trying to get random samples based on group for a relatively large data frame. I need to get unique results for each group member only - I can't have results duplicate for an individual member or overall.

I have used this code successfully for small samples:

    processors2 <- processors%>%filter(str_detect(Person.Who.Changed.Object, "A0")) %>% 
      group_by(User)%>% sample_n(., 2)

However, if I use the below similar code, I get multiple duplicates both within groups and overall (ie member 1 and member 3 get the same row of data, and member 1 gets 2 of a different row all to itself).

claimallocator2 <- claimallocator%>%
  group_by(User)%>% sample_n(80, weight = Claim.Amt)

Additionally, it makes no difference if I add replace = FALSE. I am still getting duplicates.

The expected output (obviously on a drastically smaller scale):

User    Warranty.Claim  Claim.amt
User 1  1   500
User 1  2   1000
User 1  3   1500
User 1  4   2000
User 1  5   2500
User 2  6   3000
User 2  7   3500
User 2  8   4000
User 2  9   4500
User 2  10  5000
User 2  11  5500
User 2  12  6000
User 3  13  6500
User 3  14  7000
User 3  15  7500
User 3  16  8000
User 3  17  8500
User 3  18  9000
User 3  19  9500
User 3  20  10000
User 3  21  10500
User 3  22  11000

What I am actually getting:

    User    Warranty.Claim  Claim.amt
    User 1  1   500
    User 1  1   500
    User 1  3   1500
    User 1  4   2000
    User 1  5   2500
    User 2  6   3000
    User 2  7   3500
    User 2  8   4000
    User 2  9   4500
    User 2  10  5000
    User 2  11  5500
    User 2  12  6000
    User 3  13  6500
    User 3  14  7000
    User 3  15  7500
    User 3  16  8000
    User 3  17  8500
    User 3  18  9000
    User 3  19  9500
    User 3  8   4000
    User 3  21  10500
    User 3  22  11000

Answer 1

Try this approach: first remove the duplicated rows, then group by user and sample the desired number of cases.

# create toy data
df <- data.frame(user=sample(1:10,1000,T),
                 warranty=sample(1:10,1000,T),
                 claim=sample(1:10,1000,T))

# count number of duplicate user-warranty-claim trios
df %>% count(user,warranty,claim) %>% arrange(desc(n))

# remove duplicates, sample 2 cases per user
df %>% group_by(user,warranty,claim) %>% slice(1) %>% 
  ungroup() %>% group_by(user) %>% sample_n(2)

Answer 2

您可以检查sample_n()函数中的replace选项