在R中使用across对数据框多列应用自定义函数报错求助

我有一个名为ex_ds的数据框，包含变量v1至v15：

> head(ex_ds)
  v1   v2 v3 v4 v5 v6 v7   v8 v9 v10 v11  v12  v13 v14 v15
1  5 2014  1  2  4  1  1    8  4   2   2    2    2   2   2
2  5 2014  2  6  1  3  8 <NA>  1   1   2    2    2   1   2
3  5 2014  2  5  2  1  1    8  1   1   1 <NA> <NA>   1   2
4  5 2014  2  2  1  4  1    2  5   2   2    2    2   2   2
5  5 2014  1  5  2  1  8    7  4   1   1    2    2   2   2
6  5 2014  2  4  3  5  3    1  3   2   2    2    2   2   2

我需要按v1和v2的组合分组，统计v3至v15每个变量的非NA响应数量。

我写了自定义函数，单独调用时可以正常运行：

n_fn_ex = function(var){
  n_var = ex_ds %>% 
    drop_na(var) %>%
    group_by(v1,v2) %>%
    count() 
}

单独调用结果：

> n_fn_ex("v3")
# A tibble: 1 x 3
# Groups:   v1, v2 [1]
     v1    v2     n
  <int> <int> <int>
1     5  2014     6

> n_fn_ex("v8")
# A tibble: 1 x 3
# Groups:   v1, v2 [1]
     v1    v2     n
  <int> <int> <int>
1     5  2014     5

但将其用于across语句时却报错：

ex_ds_1 = ex_ds %>%
  reframe(across(v3:v15, n_fn_ex))

错误信息：

Error in `reframe()`:
i In argument: `across(v3:v15, n_fn_ex)`.
Caused by error in `across()`:
! Can't compute column `v3`.
Caused by error in `drop_na()`:
! Can't subset columns that don't exist.
x Columns `1`, `2`, `2`, `2`, `1`, etc. don't exist.
Run `rlang::last_trace()` to see where the error occurred.

我尝试过指定字符串列名、修改函数返回值等写法，均出现相同错误。

示例数据集：

ex_ds = structure(list(v1 = c(5L, 5L, 5L, 5L, 5L, 5L), v2 = c(2014L, 
2014L, 2014L, 2014L, 2014L, 2014L), v3 = structure(c(1L, 2L, 
2L, 2L, 1L, 2L), .Label = c("1", "2"), class = "factor"), v4 = structure(c(2L, 
6L, 5L, 2L, 5L, 4L), .Label = c("1", "2", "3", "4", "5", "6"), class = "factor"), 
    v5 = structure(c(4L, 1L, 2L, 1L, 2L, 3L), .Label = c("1", 
    "2", "3", "4"), class = "factor"), v6 = structure(c(1L, 3L, 
    1L, 4L, 1L, 5L), .Label = c("1", "2", "3", "4", "5", "6"), class = "factor"), 
    v7 = structure(c(1L, 8L, 1L, 1L, 8L, 3L), .Label = c("1", 
    "2", "3", "4", "5", "6", "7", "8"), class = "factor"), v8 = structure(c(8L, 
    NA, 8L, 2L, 7L, 1L), .Label = c("1", "2", "3", "4", "5", 
    "6", "7", "8"), class = "factor"), v9 = structure(c(4L, 1L, 
    1L, 5L, 4L, 3L), .Label = c("1", "2", "3", "4", "5"), class = "factor"), 
    v10 = structure(c(2L, 1L, 1L, 2L, 1L, 2L), .Label = c("1", 
    "2"), class = "factor"), v11 = structure(c(2L, 2L, 1L, 2L, 
    1L, 2L), .Label = c("1", "2"), class = "factor"), v12 = structure(c(2L, 
    2L, NA, 2L, 2L, 2L), .Label = c("1", "2"), class = "factor"), 
    v13 = structure(c(2L, 2L, NA, 2L, 2L, 2L), .Label = c("1", 
    "2"), class = "factor"), v14 = structure(c(2L, 1L, 1L, 2L, 
    2L, 2L), .Label = c("1", "2"), class = "factor"), v15 = structure(c(2L, 
    2L, 2L, 2L, 2L, 2L), .Label = c("1", "2"), class = "factor")), row.names = c(NA, 
6L), class = "data.frame")

across在迭代列时，传递给函数的是列的数值向量，而非列名字符串。你的自定义函数n_fn_ex期望接收列名字符串，但across传入的是列的实际值（比如v3的向量c(1,2,2,2,1,2)），导致drop_na(var)尝试删除这些值对应的列，自然找不到，引发报错。另外，函数直接引用全局环境的ex_ds，没有利用across传递的分组上下文，也是问题之一。

方法1：直接在across中统计非NA数量 #

无需自定义函数，直接用sum(!is.na(.x))实现统计，简洁高效：

library(dplyr)

ex_ds_1 <- ex_ds %>%
  group_by(v1, v2) %>%
  reframe(across(v3:v15, ~sum(!is.na(.x)), .names = "n_{.col}"))

方法2：修改自定义函数适配across #

让函数接收列向量，直接统计非NA数量：

n_fn_ex_fix <- function(col_vec) {
  sum(!is.na(col_vec))
}

ex_ds_1 <- ex_ds %>%
  group_by(v1, v2) %>%
  reframe(across(v3:v15, n_fn_ex_fix, .names = "n_{.col}"))

方法3：用summarise替代reframe #

效果与上述方法一致，适合习惯用summarise的场景：

ex_ds_1 <- ex_ds %>%
  group_by(v1, v2) %>%
  summarise(across(v3:v15, ~sum(!is.na(.x))), .groups = "drop")

输出结果 #

三种方法都会得到按v1、v2分组，每个变量的非NA计数：

> ex_ds_1
# A tibble: 1 x 15
     v1    v2 n_v3 n_v4 n_v5 n_v6 n_v7 n_v8 n_v9 n_v10 n_v11 n_v12 n_v13 n_v14 n_v15
  <int> <int> <int> <int> <int> <int> <int> <int> <int>  <int>  <int>  <int>  <int>  <int>  <int>
1     5  2014     6     6     6     6     6     5     6      6      6      5      5      6      6

内容的提问来源于stack exchange，提问作者abrar