R语言基于列名模式对数据框列执行计算的实现方法

针对需要按列名匹配规则批量对列配对做计算的场景，以下两种写法比原始嵌套循环更符合R编程习惯，可读性和可维护性更高。

原始参考代码 #

df <- data.frame(xA1 = sample(x = 1:4, size = 10, replace = TRUE),
                 yA1 = sample(x = 1:4, size = 10, replace = TRUE),
                 xA2 = sample(x = 1:4, size = 10, replace = TRUE),
                 yA2 = sample(x = 1:4, size = 10, replace = TRUE),
                 xB1 = sample(x = 1:4, size = 10, replace = TRUE),
                 yB1 = sample(x = 1:4, size = 10, replace = TRUE),
                 xB2 = sample(x = 1:4, size = 10, replace = TRUE),
                 yB2 = sample(x = 1:4, size = 10, replace = TRUE))

# 原始实现：双层for循环+逐行apply
for (middle in c('A', 'B')) {
 for (right in 1:2) {
  df[paste0(middle, right)] <- apply(X = subset(df, select = c(paste0('x', middle, right), paste0('y', middle, right))),
                                     MARGIN = 1,
                                     FUN = weighted.mean,
                                     w = c(0.25, 0.75))
 }
}

优化实现方案 #

方案1：tidyverse 实现（可读性优先推荐） #

用dplyr的列选择语法直接匹配列名模式，不需要手动拼接列名写嵌套循环，管道操作逻辑清晰，后续新增列分组时不需要修改核心计算逻辑：

library(dplyr)

# 固定随机种子方便复现结果
set.seed(123)
df <- tibble(xA1 = sample(x = 1:4, size = 10, replace = TRUE),
             yA1 = sample(x = 1:4, size = 10, replace = TRUE),
             xA2 = sample(x = 1:4, size = 10, replace = TRUE),
             yA2 = sample(x = 1:4, size = 10, replace = TRUE),
             xB1 = sample(x = 1:4, size = 10, replace = TRUE),
             yB1 = sample(x = 1:4, size = 10, replace = TRUE),
             xB2 = sample(x = 1:4, size = 10, replace = TRUE),
             yB2 = sample(x = 1:4, size = 10, replace = TRUE))

df <- df %>% 
  mutate(
    across(
      .cols = starts_with("x"),
      .names = "{sub('^x', '', .col)}",
      ~ 0.25 * .x + 0.75 * cur_data()[[sub("^x", "y", cur_column())]]
    )
  )

代码逻辑说明：

• 用starts_with("x")选择器自动匹配所有x前缀的列，不需要硬编码A/B、1/2的遍历规则
• .names参数自动提取列名后缀作为新列名，不需要手动拼接字符串
• 直接用向量运算做加权计算，比逐行apply效率更高
• 后续如果新增xC1、yC1、xA3、yA3这类列，代码不需要修改就能自动计算对应新列。

方案2：base R 实现（无第三方依赖） #

如果不想加载额外包，可以用base R的向量运算实现，同样避免嵌套循环，运行效率更高：

set.seed(123)
df <- data.frame(xA1 = sample(x = 1:4, size = 10, replace = TRUE),
                 yA1 = sample(x = 1:4, size = 10, replace = TRUE),
                 xA2 = sample(x = 1:4, size = 10, replace = TRUE),
                 yA2 = sample(x = 1:4, size = 10, replace = TRUE),
                 xB1 = sample(x = 1:4, size = 10, replace = TRUE),
                 yB1 = sample(x = 1:4, size = 10, replace = TRUE),
                 xB2 = sample(x = 1:4, size = 10, replace = TRUE),
                 yB2 = sample(x = 1:4, size = 10, replace = TRUE))

# 自动提取所有x列的后缀
col_suffix <- gsub("^x", "", names(df)[startsWith(names(df), "x")])

# 批量计算加权均值
for(suf in col_suffix){
  df[[suf]] <- 0.25 * df[[paste0("x", suf)]] + 0.75 * df[[paste0("y", suf)]]
}

注意：原始代码中注释的weighted.mean(x = c(df$xA1, df$yA1), w = c(0.25, 0.75))是把两列所有值合并为一个向量计算整体加权均值，和逐行计算配对值均值的逻辑不一致，使用时需要根据实际需求选择计算逻辑。

内容的提问来源于stack exchange，提问作者user1