将满足多个条件的观察结果保留在 group_id 中

Question

我正在尝试对一些数据进行排序，这些数据标有 id、group_id 和状态代码。这是一个例子：

library(tidyverse)
sample <- tibble(id = c(1:20),
                 group_id = c(rep(1,3),rep(4,4),rep(8,5),rep(13,3),rep(16,1),rep(17,1),rep(18,3)),
                 status_code = c("initial","initial","final","initial","initial","initial","sold","initial","initial","initial","initial","error",
                            "initial","initial","final","final","error","initial","final","final")
)


sample %>%
  head()

# A tibble: 6 x 3
     id group_id status_code 
  <int>    <dbl> <chr>  
1     1        1 initial
2     2        1 initial
3     3        1 final  
4     4        4 initial
5     5        4 initial
6     6        4 initial

我想做的是按group_id 对它们进行分组，并仅保留满足以下条件的观察结果：

• 状态代码（最终、已售、错误）在group_id 中出现多次
• “final”、“sold”或“error”出现不止一次

在我上面的例子中，group_id == 18 应该被过滤，因为代码“final”出现了两次。但这也应该是“最终”代码之一被替换为“已售”或“错误”的情况。

基本上，我希望能够轻松地拆分数据，如下所示：

# Conditions met
# A tibble: 3 x 3
     id group_id status_code
  <int>    <dbl> <chr>      
1    18       18 initial    
2    19       18 final      
3    20       18 final   

# Conditions not met
# A tibble: 17 x 3
      id group_id status_code
   <int>    <dbl> <chr>      
 1     1        1 initial    
 2     2        1 initial    
 3     3        1 final      
 4     4        4 initial    
 5     5        4 initial    
 6     6        4 initial    
 7     7        4 sold       
 8     8        8 initial    
 9     9        8 initial    
10    10        8 initial    
11    11        8 initial    
12    12        8 error      
13    13       13 initial    
14    14       13 initial    
15    15       13 final      
16    16       16 final      
17    17       17 error    

我的数据显然有更多的列，应该都保留在每个子帧中。到目前为止，我已经能够通过计数、保留 status_code 并再次过滤主要数据来实现这一点：

condition <- sample %>%
  filter(status_code != "initial") %>% # Removing "initial" status_code
  count(group_id) %>% # Counting the number of remaining occurences in each group
  filter(n > 1) %>% # Keeping groups where more than one occurence
  select(group_id) # Selecting the group_ids

fitlered_sample <- left_join(condition,sample, by ="group_id")

# A tibble: 3 x 3
  group_id    id status_code
     <dbl> <int> <chr>      
1       18    18 initial    
2       18    19 final      
3       18    20 final   

它有效，但我希望这可以在不创建“条件”数据框的情况下实现。我更喜欢 dplyr 解决方案，但我也可以使用 base 或 data.table

Answer 1

我们可以通过 'group_id' 分组，然后 filter 计数 'status_code' 不等于 'initial'

library(dplyr)
sample %>% 
 group_by(group_id) %>%
 filter(sum(status_code != "initial") > 1) %>%
 ungroup

其他数据

sample %>% 
  group_by(group_id) %>% 
  filter(sum(status_code != "initial") == 1) %>%
  ungroup

---

或者使用group_split创建一个list的数据集

lst1 <- sample %>% 
 group_by(group_id) %>% 
 mutate(n = sum(status_code != "initial") != 1) %>% 
 ungroup %>% 
 group_split(n, .keep = FALSE)

-输出

lst1
[[1]]
# A tibble: 17 × 3
      id group_id status_code
   <int>    <dbl> <chr>      
 1     1        1 initial    
 2     2        1 initial    
 3     3        1 final      
 4     4        4 initial    
 5     5        4 initial    
 6     6        4 initial    
 7     7        4 sold       
 8     8        8 initial    
 9     9        8 initial    
10    10        8 initial    
11    11        8 initial    
12    12        8 error      
13    13       13 initial    
14    14       13 initial    
15    15       13 final      
16    16       16 final      
17    17       17 error      

[[2]]
# A tibble: 3 × 3
     id group_id status_code
  <int>    <dbl> <chr>      
1    18       18 initial    
2    19       18 final      
3    20       18 final      

Answer 2

与data.table:

library(data.table)
setDT(sample)

sample[,.SD[sum(status_code=="final")>1|
            sum(status_code=="sold") >1|
            sum(status_code=="error")>1],
       by=group_id]

   group_id    id status_code
      <num> <int>      <char>
1:       18    18     initial
2:       18    19       final
3:       18    20       final

相反的：

sample[,.SD[!(sum(status_code=="final")>1|
              sum(status_code=="sold") >1|
              sum(status_code=="error")>1)],
       by=group_id]


    group_id    id status_code
       <num> <int>      <char>
 1:        1     1     initial
 2:        1     2     initial
 3:        1     3       final
 4:        4     4     initial
 5:        4     5     initial
 6:        4     6     initial
 7:        4     7        sold
 8:        8     8     initial
 9:        8     9     initial
10:        8    10     initial
11:        8    11     initial
12:        8    12       error
13:       13    13     initial
14:       13    14     initial
15:       13    15       final
16:       16    16       final
17:       17    17       error