按行检查NA，然后汇总其列名

如何解决按行检查NA，然后汇总其列名

我目前正在努力逐行获取信息，这些信息的变数很好/为NA。下面，我显示了一部分数据的示例/输出输出。

head(big_test)
# A tibble: 3 x 19
     id ctr_n   ctr    yr    mn  nvvi ENP_nat ENP_avg ENP_wght inflation1 inflation2 inflation3 inflation4 PSNS  PSNS_s PSNS_w
  <dbl> <chr> <dbl> <dbl> <dbl> <dbl>   <dbl>   <dbl>    <dbl> <chr>      <chr>      <chr>      <chr>      <chr>  <dbl> <chr> 
1  1854 Isra~   376  2019     4     1    3.50    3.50     3.50 NA         NA         NA         NA         NA     0.962 NA    
2  1855 Isra~   376  2019     9     1    2.51    2.51     2.51 NA         NA         NA         NA         NA     0.992 NA    
3  1856 Isra~   376  2020     3     1    3.78    3.78     3.78 NA         NA         NA         NA         NA     0.999 NA    
# ... with 3 more variables: PSNS_sw <chr>,local_E <dbl>,cst_tot <dbl>

dput(big_test)
structure(list(id = c(1854,1855,1856),ctr_n = c("Israel","Israel","Israel"),ctr = c(376,376,376),yr = c(2019,2019,2020),mn = c(4,9,3),nvvi = c(1,1,1),ENP_nat = c(3.50348063163162,2.51319610127466,3.78468892335972),ENP_avg = c(3.50348063163162,ENP_wght = c(3.50348063163162,inflation1 = c("NA","NA","NA"),inflation2 = c("NA",inflation3 = c("NA",inflation4 = c("NA",PSNS = c("NA",PSNS_s = c(0.961748183147869,0.992275075925835,0.998547438416594),PSNS_w = c("NA",PSNS_sw = c("NA",local_E = c(1,cst_tot = c(1,1)),row.names = c(NA,-3L),class = c("tbl_df","tbl","data.frame"))

编辑：此处的NA用引号引起来，这是不合适的。我认为问题出在写.xlsx。正确的版本在下面的dput输出中列出，NA上没有引号。

如您所见，数据在全国范围内通过选举分开，此处的每一行都应该是唯一的（即以色列，2019年，第4个月）。 我要创建一个字符列，其中列出了此输出中缺少的变量。这是所需列的示例：

desired_output <- tibble(missing_vars=paste("inflation1","inflation2","inflation3","inflation4","etc",sep=";"))

head(desired_output)
# A tibble: 1 x 1
  missing_vars                                   
  <chr>                                          
1 inflation1;inflation2;inflation3;inflation4;etc

因此，我想知道是否存在某种循环，可以对唯一的选举进行切片，查看丢失的列，然后对丢失的列进行矢量处理？这对于自动化至关重要，因为在存在相同国家/地区的其他变量的情况下，某些变量可能会丢失。我曾尝试对它们进行计数，但无法弄清楚如何将这些列名称列出为字符列。

感谢您的帮助。谢谢！

解决方法

这是tidyverse的一个选项，在这里我们用pivot_longer重塑为'long'格式，并按row_number()，paste分组，其中缺少的列名“值”中的值

library(dplyr)
library(tidyr)
library(stringr)
big_test %>%
     select(starts_with('inflation')) %>% 
     mutate(rn = row_number()) %>% 
     pivot_longer(cols = -rn) %>% 
     group_by(rn) %>%
     summarise(missing_vars = str_c(name[is.na(value)],collapse=";"),.groups = 'drop') %>%
     select(-rn)          

---

在不重塑的情况下，选项为rowwise/c_across

big_test %>% 
   rowwise %>% 
   transmute(missing_vars = str_c(names(select(cur_data(),starts_with('inflation')))[which(c_across(starts_with('inflation')) 
        == 'NA')],collapse=";"))

在这里，它将（==）与“ NA”进行比较。如果它是真实的NA，请使用is.na代替==

,

@akrun指出，您使用的是“ NA”字符串而不是NA。解决该问题后，您可以定义一个缺少的函数并将其应用于每一行以创建一个新变量：

big_test <- structure(list(id = c(1854,1855,1856),ctr_n = c("Israel","Israel","Israel"),ctr = c(376,376,376),yr = c(2019,2019,2020),mn = c(4,9,3),nvvi = c(1,1,1),ENP_nat = c(3.50348063163162,2.51319610127466,3.78468892335972),ENP_avg = c(3.50348063163162,ENP_wght = c(3.50348063163162,inflation1 = c(NA,NA,NA),inflation2 = c(NA,inflation3 = c(NA,inflation4 = c(NA,PSNS = c(NA,PSNS_s = c(0.961748183147869,0.992275075925835,0.998547438416594),PSNS_w = c(NA,PSNS_sw = c(NA,local_E = c(1,cst_tot = c(1,1)),row.names = c(NA,-3L),class = c("tbl_df","tbl","data.frame"))

missing <- function(x) {
  idx <- is.na(unlist(x))
  paste(colnames(big_test)[idx],collapse=",")
}

big_test$missing <- apply(big_test,missing)

big_test$missing
#> [1] "inflation1,inflation2,inflation3,inflation4,PSNS,PSNS_w,PSNS_sw"
#> [2] "inflation1,PSNS_sw"
#> [3] "inflation1,PSNS_sw"