在R中仅针对确诊组删除含超过2个NA的行

原始数据定义 #

ID <- c("x", "y", "z", "ad", "tgfg", "gfgh", "asj", "gh", "sdf", "asdgz")
diagnosis <- c("yes", "no", "yes", "yes", "yes", "no", "yes", "no", "no", "no")
Q1 <- c("A",NA,"A",NA,NA,"C","D","A","B", NA)
Q2 <- c("D",NA,"D","C",NA,NA,"A","A","A","A")
Q3 <- c("B","B","C","A",NA,"A","B","D","E",NA)
Q4 <- c("B",NA,"C","C","C","C","D","B",NA,"A")
mydf <- data.frame(ID, diagnosis, Q1,Q2,Q3,Q4)

数据展示 #

ID diagnosis   Q1   Q2   Q3   Q4
x       yes    A    D    B    B
y        no <NA> <NA>    B <NA>
z       yes    A    D    C    C
ad      yes <NA>    C    A    C
tgfg    yes <NA> <NA> <NA>    C
gfgh     no    C <NA>    A    C
asj     yes    D    A    B    D
gh       no    A    A    D    B
sdf      no    B    A    E <NA>
asdgz    no <NA>    A <NA>    A

需求说明 #

仅删除diagnosis为"yes"组中含超过2个NA的行，diagnosis为"no"的组无论NA数量多少都保留。

你尝试的错误代码 #

delete.na <- function(DF, n=0) {
  DF[rowSums(is.na(DF)) <= n,]
}
new_df <- mydf %>% group_by(diagnosis== "yes") %>% delete.na(2) %>% ungroup()

错误原因 #

你用group_by(diagnosis== "yes")会把数据分成**TRUE（yes组）和FALSE（no组）**两组，然后对两组都执行了delete.na(2)过滤，导致"no"组里NA数超过2的行（比如y行有3个NA）被误删。

正确解决方案 #

方法1：直接用dplyr的filter（推荐）

不需要分组，直接通过条件筛选：要么是"no"组，要么是"yes"组且NA数≤2。

library(dplyr)

new_df <- mydf %>%
  # 计算Q1-Q4列的NA数量
  mutate(na_count = rowSums(is.na(select(., Q1:Q4)))) %>%
  # 核心筛选逻辑
  filter(diagnosis == "no" | (diagnosis == "yes" & na_count <= 2)) %>%
  # 移除临时计算的na_count列
  select(-na_count)

方法2：分组后针对性处理

如果想用分组逻辑，可以按diagnosis分组，对不同组应用不同规则：

delete.na <- function(DF, n=0) {
  DF[rowSums(is.na(DF)) <= n,]
}

new_df <- mydf %>%
  group_by(diagnosis) %>%
  group_modify(function(data, key) {
    if(key$diagnosis == "yes") {
      delete.na(data, 2) # yes组执行NA过滤
    } else {
      data # no组直接保留原数据
    }
  }) %>%
  ungroup()

执行后，"no"组的y行会被完整保留，"yes"组中只有tgfg行（含3个NA）会被删除，完全符合需求。

内容的提问来源于stack exchange，提问作者dplyr