R rowSums in case_when

Question

考虑到这是我的数据集

df <- 
structure(list(id = 1:6, w = c(NA, 11L, 12L, 13L, 14L, 15L), 
    x = c(20L, 21L, NA, 23L, 24L, 25L), y = c(30L, 31L, 32L, 
    NA, 34L, 35L), z = c(40L, NA, 42L, 43L, 44L, NA), Group = c("Group1", 
    "Group2", "Group2", "Group2", "Group3", "Group3")), row.names = c(NA, 
-6L), class = c("tbl_df", "tbl", "data.frame"))


    id     w     x     y     z Group   
     1    NA    20    30    40 Group1
     2    11    21    31    NA Group2
     3    12    NA    32    42 Group2
     4    13    23    NA    43 Group2
     5    14    24    34    44 Group3
     6    15    25    35    NA Group3

我试图从w，x，y，z列中按组对观察结果进行求和。我使用这样的case_when语句。

df <- df %>%
  mutate(
    
    group1_total <- case_when( Group = "Group1" ~ rowSums(w,x, na.rm = TRUE)),
    group2_total <- case_when( Group = "Group2" ~ rowSums(w,x,z, na.rm = TRUE)),
    group3_total <- case_when( Group = "Group3" ~ rowSums(w,x,y,z, na.rm = TRUE))   
    
  )

我收到一个错误的说法，rowSums()中的错误。不知道我做错了什么。任何建议或帮助都是非常感谢的，谢谢。

Answer 1

我将创建一个命名列表，在该列表中，我们将为每个组指定要求和的变量。

cur_data()返回当前组的数据，group_cols[[unique(Group)]]为每个组选择适当的列。

library(dplyr)

group_cols <- list('Group1' = c('w', 'x'), 'Group2' = c('w', 'x', 'z'), 
                   'Group3' = c('w', 'x', 'y', 'z'))

df %>%
  group_by(Group) %>%
  mutate(total = rowSums(select(cur_data(), 
                         group_cols[[unique(Group)]]), na.rm = TRUE)) %>%
  ungroup

#     id     w     x     y     z Group  total
#  <int> <int> <int> <int> <int> <chr>  <dbl>
#1     1    NA    20    30    40 Group1    20
#2     2    11    21    31    NA Group2    32
#3     3    12    NA    32    42 Group2    54
#4     4    13    23    NA    43 Group2    79
#5     5    14    24    34    44 Group3   116
#6     6    15    25    35    NA Group3    75

这样做的好处是，计算是按组进行的，而不是每一行。

Answer 2

您可以将rowwise()、c_across() (c_across()设计为与rowwise()一起使用以便于执行逐行聚合)和sum()结合使用。

library(dplyr)

df %>%
 rowwise() %>%   
  mutate(
    group1_total = case_when( Group == "Group1" ~ sum(c_across(w:x), na.rm = TRUE)),
    group2_total = case_when( Group == "Group2" ~ sum(c_across(c(w,x,z)), na.rm = TRUE)),
    group3_total = case_when( Group == "Group3" ~ sum(c_across(w:z), na.rm = TRUE))   
  )

输出

     id     w     x     y     z Group  group1_total group2_total group3_total
  <int> <int> <int> <int> <int> <chr>         <int>        <int>        <int>
1     1    NA    20    30    40 Group1           20           NA           NA
2     2    11    21    31    NA Group2           NA           32           NA
3     3    12    NA    32    42 Group2           NA           54           NA
4     4    13    23    NA    43 Group2           NA           79           NA
5     5    14    24    34    44 Group3           NA           NA          116
6     6    15    25    35    NA Group3           NA           NA           75

如果您对一个总列感兴趣，那么可以使用coalesce()，例如。

df %>%
  rowwise() %>%   
  mutate(
    group1_total = case_when( Group == "Group1" ~ sum(c_across(w:x), na.rm = TRUE)),
    group2_total = case_when( Group == "Group2" ~ sum(c_across(c(w,x,z)), na.rm = TRUE)),
    group3_total = case_when( Group == "Group3" ~ sum(c_across(w:z), na.rm = TRUE))   
  ) %>% 
  mutate(total = coalesce(group1_total, group2_total,group3_total)) %>% 
  select(-contains("_total"))

导致

# A tibble: 6 x 7
# Rowwise: 
     id     w     x     y     z Group  total
  <int> <int> <int> <int> <int> <chr>  <int>
1     1    NA    20    30    40 Group1    20
2     2    11    21    31    NA Group2    32
3     3    12    NA    32    42 Group2    54
4     4    13    23    NA    43 Group2    79
5     5    14    24    34    44 Group3   116
6     6    15    25    35    NA Group3    75

Answer 3

一个选项是按键/值数据集连接，替换与NA不匹配的列，然后获取rowSums

library(dplyr)
library(stringr)
keydat <- tibble(Group = c("Group1", "Group2", "Group3"), 
   nm = c("w|x",  "w|x|z", 'w|x|y|z'))
df %>% 
  left_join(keydat) %>%
   mutate(total = rowSums(across(w:z, ~ case_when(str_detect(cur_column(),
       nm)~ .x)), na.rm = TRUE), nm = NULL)

-output

# A tibble: 6 × 7
     id     w     x     y     z Group  total
  <int> <int> <int> <int> <int> <chr>  <dbl>
1     1    NA    20    30    40 Group1    20
2     2    11    21    31    NA Group2    32
3     3    12    NA    32    42 Group2    54
4     4    13    23    NA    43 Group2    79
5     5    14    24    34    44 Group3   116
6     6    15    25    35    NA Group3    75