group_by，查找组的值，返回值和原始data.frame？ （有效地，使用 dplyr）

Question

我正在尝试group_by，找到符合条件的行的第一次出现，并创建一个新列，其值基于从每个组中选择的行。

示例

这很容易用一个例子来演示。一个新列应该由

生成

1. 按 transaction_id 分组
2. 寻找第一次出现的icecream_bool (= 1)（注意第二笔交易有两行符合这个条件，所以应该取第一笔）
3. 使用“item”列中的值创建新列

我们从这个data.frame开始

df <- data.frame(
  transaction_id=as.integer(c(1,1,1,2,2,2,2,2,3,3,3)),
  item=as.character(c("crisps", "magnum", "gum",
                      "jerky", "cheese", "snickers", "ben&jerry", "magnum",
                      "halo", "crisps", "mars")),
  icecream_bool=as.integer(c(0,1,0,
                             0,0,0,1,1,
                             1,0,0)),
  stringsAsFactors = F
)

#    transaction_id      item icecream_bool
# 1               1    crisps             0
# 2               1    magnum             1
# 3               1       gum             0
# 4               2     jerky             0
# 5               2    cheese             0
# 6               2  snickers             0
# 7               2 ben&jerry             1
# 8               2    magnum             1
# 9               3      halo             1
# 10              3    crisps             0
# 11              3      mars             0

期望的输出

像这样生成 ice_cream 列

   transaction_id      item icecream_bool ice_cream
1               1    crisps             0    magnum
2               1    magnum             1    magnum
3               1       gum             0    magnum
4               2     jerky             0 ben&jerry
5               2    cheese             0 ben&jerry
6               2  snickers             0 ben&jerry
7               2 ben&jerry             1 ben&jerry
8               2    magnum             1 ben&jerry
9               3      halo             1      halo
10              3    crisps             0      halo
11              3      mars             0      halo

Answer 1

如果icecream_bool 中只有 1 和 0，另一种方法是使用 which.max

library(dplyr)
df %>%
  group_by(transaction_id) %>%
  mutate(ice_cream = item[which.max(icecream_bool)])


#   transaction_id item      icecream_bool ice-cream
#            <int> <chr>             <int> <chr>    
# 1              1 crisps                0 magnum   
# 2              1 magnum                1 magnum   
# 3              1 gum                   0 magnum   
# 4              2 jerky                 0 ben&jerry
# 5              2 cheese                0 ben&jerry
# 6              2 snickers              0 ben&jerry
# 7              2 ben&jerry             1 ben&jerry
# 8              2 magnum                1 ben&jerry
# 9              3 halo                  1 halo     
#10              3 crisps                0 halo     
#11              3 mars                  0 halo     

这个也可以写成data.table

library(data.table)
setDT(df)[, ice_cream := item[which.max(icecream_bool)], transaction_id]

---

如果有大于 1 的值，我们需要与 1 进行比较，我们可以使用match

df %>%
  group_by(transaction_id) %>%
  mutate(ice_cream = item[match(1,icecream_bool)])

Answer 2

使用 dplyr 的一种方式

library(dplyr)
df%>%
  group_by(transaction_id, icecream_bool) %>%
  mutate(mr = row_number()) %>%
  group_by(transaction_id) %>%
  mutate(ice_cream=item[mr==1 & icecream_bool==1]) %>%
  select(-mr)

---

# A tibble: 11 x 4
# Groups:   transaction_id [3]
   transaction_id item      icecream_bool ice_cream
            <int> <chr>             <int> <chr>    
 1              1 crisps                0 magnum   
 2              1 magnum                1 magnum   
 3              1 gum                   0 magnum   
 4              2 jerky                 0 ben&jerry
 5              2 cheese                0 ben&jerry
 6              2 snickers              0 ben&jerry
 7              2 ben&jerry             1 ben&jerry
 8              2 magnum                1 ben&jerry
 9              3 halo                  1 halo     
10              3 crisps                0 halo     
11              3 mars                  0 halo 

Answer 3

你可以这样做：

library(dplyr)

df %>%
  group_by(transaction_id) %>%
  mutate(icecream = first(item[icecream_bool == 1]))

# A tibble: 11 x 4
# Groups:   transaction_id [3]
   transaction_id item      icecream_bool ice      
            <int> <chr>             <int> <chr>    
 1              1 crisps                0 magnum   
 2              1 magnum                1 magnum   
 3              1 gum                   0 magnum   
 4              2 jerky                 0 ben&jerry
 5              2 cheese                0 ben&jerry
 6              2 snickers              0 ben&jerry
 7              2 ben&jerry             1 ben&jerry
 8              2 magnum                1 ben&jerry
 9              3 halo                  1 halo     
10              3 crisps                0 halo     
11              3 mars                  0 halo   

Answer 4

我不确定它是否有效，但我找到了一种方法

df <- data.frame(
  transaction_id=as.integer(c(1,1,1,2,2,2,2,2,3,3,3)),
  item=as.character(c("crisps", "magnum", "gum",
                      "jerky", "cheese", "snickers", "ben&jerry", "magnum",
                      "halo", "crisps", "mars")),
  icecream_bool=as.integer(c(0,1,0,
                             0,0,0,1,1,
                             1,0,0)),
  stringsAsFactors = F
)

df %>% 
  group_by(transaction_id) %>% 
  filter(icecream_bool == 1) %>% 
  distinct(transaction_id, icecream_bool, .keep_all = T) %>% 
  select(-icecream_bool) %>% 
  rename(icecream=item) %>% 
  left_join(df, ., by=c("transaction_id"="transaction_id"))


   transaction_id      item icecream_bool  icecream
1               1    crisps             0    magnum
2               1    magnum             1    magnum
3               1       gum             0    magnum
4               2     jerky             0 ben&jerry
5               2    cheese             0 ben&jerry
6               2  snickers             0 ben&jerry
7               2 ben&jerry             1 ben&jerry
8               2    magnum             1 ben&jerry
9               3      halo             1      halo
10              3    crisps             0      halo
11              3      mars             0      halo