R语言中简化多方法排名生成与Spearman相关性计算的方法

需求说明 #

原代码通过4条独立命令生成各方法的排名，再筛选列计算排名间的Spearman系数，希望实现更简洁的批量处理流程。

优化后的代码 #

library(dplyr)

# 原始数据
padr <- structure(list(Method1= c(0.343394182514031, 1, 0.860087696840587, 
0.860087696840587, 0.868085451239441, 0.698055447477473, 0.43737803420133, 
0.434970400304271, 0.434970400304271, 0.379233994071699), Method2 = c(1, 
0.232979733215734, 0.240392548713602, 0.240392548713602, 0.213384133751235, 
0.240137915565427, 0.321393780370283, 0.322481353908317, 0.322481353908317, 
0.352233249467427), Method3 = c(1, 0.214432400448801, 0.214809476505306, 
0.214809476505306, 0.16783443847331, 0.210797750473198, 0.293103343189013, 
0.293692283587016, 0.293692283587016, 0.281085590908947), Method4 = c(0, 
1, 0.875556823046433, 0.875556823046433, 0.891768819029077, 0.832271929255291, 
0.741168314099481, 0.740578512687553, 0.740578512687553, 0.819053554576837
)), class = "data.frame", row.names = c("1", "2", "3", "4", "5", 
"6", "7", "8", "9", "10"))

# 链式操作完成排名生成与Spearman系数计算
padr %>%
  # 批量生成排名列，按需求指定不同的ties.method
  mutate(
    RankMethod1 = (n() + 1) - rank(Method1, ties.method = "last"),
    across(c(Method2, Method3), ~(n() + 1) - rank(., ties.method = "first"), .names = "Rank{.col}"),
    RankMethod4 = (n() + 1) - rank(Method4, ties.method = "last")
  ) %>%
  # 自动匹配所有排名列，计算与RankMethod1的Spearman系数
  summarise(across(starts_with("Rank"), ~cor.test(., RankMethod1, method = "spearman")$estimate))

优化点说明 #

• 用across批量处理Method2、Method3的排名生成，减少重复代码
• 用starts_with("Rank")自动匹配所有排名列，无需手动指定列索引或范围
• 全程链式操作，无需中间筛选列的步骤，代码更简洁流畅

输出结果 #

RankMethod1 RankMethod2 RankMethod3 RankMethod4
1           1  -0.9515152   -0.830303   0.9272727

内容的提问来源于stack exchange，提问作者Antonio