在R语言中按条件替换数据框中的NA为指定字符串

我有如下所示的R数据框：

df <- as.data.frame(matrix(c("True Organic", "True Organic", NA, NA, NA, 0, 
                             "True Organic", "True Organic", NA, NA, NA, 0, 
                             "Organic Search (SEO)", "Induced Organic", NA, NA, NA, 0, 
                             "Display", NA, NA, NA, NA, 0, 
                             "Social Ads (Act)", "Induced Organic", "Induced Organic", NA, NA, 1, 
                             "Referral", "Social Ads (Act)", NA, NA, NA, 0, 
                             "Special Emails", "Induced Organic", NA, NA, NA, 1, 
                             "Daily Email", "Daily Email", "Daily Email", NA, NA, 0), 
                           nrow = 8, ncol = 6, byrow = TRUE, 
                           dimnames = list(NULL, c("Node_1", "Node_2", "Node_3", "Node_4", "Node_5", "conversion"))), 
                    stringsAsFactors = FALSE)

原始数据框输出：

Node_1                Node_2                Node_3 Node_4 Node_5 conversion
1 True Organic         True Organic         <NA>   <NA>   <NA>   0         
2 True Organic         True Organic         <NA>   <NA>   <NA>   0         
3 Organic Search (SEO) Induced Organic      <NA>   <NA>   <NA>   0         
4 Display              <NA>                 <NA>   <NA>   <NA>   0         
5 Social Ads (Act)     Induced Organic      Induced Organic <NA>   <NA>   1         
6 Referral             Social Ads (Act)     <NA>   <NA>   <NA>   0         
7 Special Emails       Induced Organic      <NA>   <NA>   <NA>   1         
8 Daily Email          Daily Email          Daily Email <NA>   <NA>   0         

我的需求是：对每行进行操作，若该行conversion列值为0，则将该行所有NA替换为字符串"Null"；若conversion列值为1，则替换为"Conversion"。期望得到的最终输出如下：

df_desired <- as.data.frame(matrix(c("True Organic", "True Organic", "Null", "Null", "Null", 0, 
                                     "True Organic", "True Organic", "Null", "Null", "Null", 0, 
                                     "Organic Search (SEO)", "Induced Organic", "Null", "Null", "Null", 0, 
                                     "Display", "Null", "Null", "Null", "Null", 0, 
                                     "Social Ads (Act)", "Induced Organic", "Induced Organic", "Conversion", "Conversion", 1, 
                                     "Referral", "Social Ads (Act)","Null", "Null", "Null", 0, 
                                     "Special Emails", "Induced Organic", "Conversion", "Conversion", "Conversion", 1, 
                                     "Daily Email", "Daily Email", "Daily Email", "Null", "Null", 0), 
                                   nrow = 8, ncol = 6, byrow = TRUE, 
                                   dimnames = list(NULL, c("Node_1", "Node_2", "Node_3", "Node_4", "Node_5", "conversion"))), 
                            stringsAsFactors = FALSE)

目标数据框输出：

Node_1                Node_2                Node_3          Node_4          Node_5          conversion
1 True Organic         True Organic         Null            Null            Null            0         
2 True Organic         True Organic         Null            Null            Null            0         
3 Organic Search (SEO) Induced Organic      Null            Null            Null            0         
4 Display              Null                 Null            Null            Null            0         
5 Social Ads (Act)     Induced Organic      Induced Organic Conversion     Conversion     1         
6 Referral             Social Ads (Act)     Null            Null            Null            0         
7 Special Emails       Induced Organic      Conversion     Conversion     Conversion     1         
8 Daily Email          Daily Email          Daily Email     Null            Null            0         

我目前可以通过嵌套for循环实现该功能：

for (i in 1:nrow(df)){ 
  for (j in 1:ncol(df)){ 
    df[i,j] <- ifelse(((is.na(df[i,j])) & df[i,]$conversion == "1"), "Conversion", df[i,j]) 
    for (j in 1:ncol(df)){ 
      df[i,j] <- ifelse(((is.na(df[i,j])) & df[i,]$conversion == "0"), "Null", df[i,j]) 
    } 
  } 
}

但该方法扩展性不佳，请问有更优的实现方案吗？感谢您的建议！

更优的实现方案 #

在R里，嵌套循环确实容易写得冗余，而且数据量大的时候运行效率很低，咱们可以用更简洁高效的方法来实现需求，下面给你几个不同的方案，按需选择就行：

方案1：用dplyr的行级操作（可读性强）

如果你熟悉tidyverse语法，dplyr的行操作非常直观，结合across()可以批量处理多列：
首先确保安装并加载dplyr：

install.packages("dplyr")
library(dplyr)

然后执行替换：

df_optimized <- df %>%
  rowwise() %>%
  mutate(across(Node_1:Node_5, ~ ifelse(is.na(.), 
                                         ifelse(conversion == "1", "Conversion", "Null"), 
                                         .))) %>%
  ungroup()

方案2：基础R的apply()函数（无需额外包）

用基础R的apply()按行处理，避免嵌套循环的冗余：

df_optimized <- as.data.frame(t(apply(df, 1, function(row) {
  na_pos <- is.na(row)
  if (row["conversion"] == "1") {
    row[na_pos] <- "Conversion"
  } else {
    row[na_pos] <- "Null"
  }
  row
})), stringsAsFactors = FALSE)

# 恢复列名和数据类型
colnames(df_optimized) <- colnames(df)
df_optimized$conversion <- as.character(df_optimized$conversion)

方案3：完全向量化操作（效率最高）

这是R里处理这类批量替换问题的首选方式，完全利用向量化特性，不用逐行遍历，速度最快代码也最简洁：

# 复制原始数据框避免修改原数据
df_optimized <- df

# 生成NA位置的掩码
na_mask <- is.na(df_optimized)

# 根据conversion的值批量替换NA
df_optimized[na_mask & df_optimized$conversion == "1"] <- "Conversion"
df_optimized[na_mask & df_optimized$conversion == "0"] <- "Null"

方案对比 #

• 嵌套循环：代码冗余，效率低，数据量大时性能下降明显，扩展性差。
• dplyr行操作：可读性强，适合习惯tidyverse的用户，处理中等规模数据友好。
• apply函数：基础R原生实现，无需额外安装包，代码简洁。
• 向量化操作：执行效率最高，代码最简洁，适合大规模数据处理，是最优选择。

内容的提问来源于stack exchange，提问作者jalind