purrr -- 将特定模式中的所有变量循环到运行相关分析

Question

假设我有这样的数据集：

ds <- data.frame(question1 = sample(1:10),
                 question2 = sample(1:10),
                 question3 = sample(1:10),

                 question1_a = sample(1:10),
                 question2_a = sample(1:10),
                 question3_a = sample(1:10))

我想检查问题 1 和 question1_a 的类内相关性；问题 2 和 question2_a；问题 3 和 question3_a。 [就是这个模式=数字+number_a].

我要走这条路：

library(psych)
library(tidyverse)

ICC(cbind(ds$question1, ds$question1_a))
ICC(cbind(ds$question2, ds$question2_a))

但我很确定可以在 tidyverse 函数 (purrr / map) 中执行所有 ICC 分析，但我很难实施此解决方案。

感谢您的帮助。

完整代码：

library(psych)
library(tidyverse)
ds <- data.frame(question1 = sample(1:10),
                 question2 = sample(1:10),
                 question3 = sample(1:10),

                 question1_a = sample(1:10),
                 question2_a = sample(1:10),
                 question3_a = sample(1:10))

ICC(cbind(ds$question1, ds$question1_a))
ICC(cbind(ds$question2, ds$question2_a))

Answer 1

我们可以这样实现：

 set.seed(233)  
  library(dplyr)
  # There might be a better way rather than selecting twice.
  # assumes an ordered dataframe
    ds %>% 
      select(matches(".*\d$")) -> df1
    ds %>% 
      select(contains("_")) -> df2 


Map(function(x, y) psych::ICC(cbind(x,y)), df1, df2)

或仅使用 base。我确实注意到，由于重复的子集和对 cbind 的进一步调用，这可能在计算上很昂贵。

Map(function(x, y) psych::ICC(cbind(x,y)), ds[grep(".*\d$",names(ds))], 
    ds[grep("_",names(ds))])

示例输出：

$question1
Call: psych::ICC(x = cbind(x, y))

Intraclass correlation coefficients 
                         type  ICC   F df1 df2    p lower bound upper bound
Single_raters_absolute   ICC1 0.24 1.6   9  10 0.23       -0.40        0.73
Single_random_raters     ICC2 0.24 1.6   9   9 0.24       -0.41        0.73
Single_fixed_raters      ICC3 0.24 1.6   9   9 0.24       -0.42        0.73
Average_raters_absolute ICC1k 0.38 1.6   9  10 0.23       -1.33        0.84
Average_random_raters   ICC2k 0.38 1.6   9   9 0.24       -1.38        0.85
Average_fixed_raters    ICC3k 0.38 1.6   9   9 0.24       -1.48        0.85

 Number of subjects = 10     Number of Judges =  2
$question2
Call: psych::ICC(x = cbind(x, y))

数据（种子设置为233）：

ds <- structure(list(question1 = c(6L, 8L, 5L, 9L, 3L, 10L, 4L, 2L, 
7L, 1L), question2 = c(7L, 2L, 3L, 10L, 8L, 9L, 6L, 5L, 1L, 4L
), question3 = c(3L, 1L, 2L, 9L, 6L, 8L, 7L, 5L, 10L, 4L), question1_a = c(4L, 
6L, 2L, 8L, 10L, 9L, 5L, 7L, 1L, 3L), question2_a = c(9L, 1L, 
10L, 5L, 4L, 2L, 8L, 7L, 3L, 6L), question3_a = c(9L, 4L, 3L, 
6L, 1L, 5L, 10L, 2L, 8L, 7L)), class = "data.frame", row.names = c(NA, 
-10L))

Answer 2

这是一个使用 purrr 且无需创建中间数据帧的解决方案

library(psych)
library(tidyverse)

ds <- data.frame(question1 = sample(1:10),
                 question2 = sample(1:10),
                 question3 = sample(1:10),

                 question1_a = sample(1:10),
                 question2_a = sample(1:10),
                 question3_a = sample(1:10))

map2(ds %>% select(matches(".*\d$")), ds %>% select(matches(".*_a$")), 
     function (x, y) ICC(cbind(x, y)))

这两个正则表达式匹配以数字结尾的列和以“_a”结尾的列。它产生

$question1
Call: ICC(x = cbind(x, y))

Intraclass correlation coefficients 
                         type  ICC   F df1 df2     p lower bound upper bound
Single_raters_absolute   ICC1 0.52 3.2   9  10 0.044      -0.090        0.85
Single_random_raters     ICC2 0.52 3.2   9   9 0.051      -0.094        0.85
Single_fixed_raters      ICC3 0.52 3.2   9   9 0.051      -0.122        0.85
Average_raters_absolute ICC1k 0.68 3.2   9  10 0.044      -0.199        0.92
Average_random_raters   ICC2k 0.68 3.2   9   9 0.051      -0.206        0.92
Average_fixed_raters    ICC3k 0.68 3.2   9   9 0.051      -0.277        0.92

 Number of subjects = 10     Number of Judges =  2
$question2
Call: ICC(x = cbind(x, y))

Intraclass correlation coefficients 
                         type  ICC   F df1 df2    p lower bound upper bound
Single_raters_absolute   ICC1 0.37 2.2   9  10 0.12       -0.27        0.79
Single_random_raters     ICC2 0.37 2.2   9   9 0.13       -0.28        0.79
Single_fixed_raters      ICC3 0.37 2.2   9   9 0.13       -0.30        0.79
Average_raters_absolute ICC1k 0.54 2.2   9  10 0.12       -0.75        0.88
Average_random_raters   ICC2k 0.54 2.2   9   9 0.13       -0.77        0.88
Average_fixed_raters    ICC3k 0.54 2.2   9   9 0.13       -0.86        0.89

 Number of subjects = 10     Number of Judges =  2
$question3
Call: ICC(x = cbind(x, y))

Intraclass correlation coefficients 
                         type  ICC   F df1 df2     p lower bound upper bound
Single_raters_absolute   ICC1 0.45 2.6   9  10 0.074       -0.18        0.82
Single_random_raters     ICC2 0.45 2.6   9   9 0.083       -0.18        0.83
Single_fixed_raters      ICC3 0.45 2.6   9   9 0.083       -0.21        0.83
Average_raters_absolute ICC1k 0.62 2.6   9  10 0.074       -0.44        0.90
Average_random_raters   ICC2k 0.62 2.6   9   9 0.083       -0.45        0.90
Average_fixed_raters    ICC3k 0.62 2.6   9   9 0.083       -0.53        0.91

 Number of subjects = 10     Number of Judges =  2

purrr -- 将特定模式中的所有变量循环到 运行 相关分析

purrr -- looping all variables within a specific pattern to run correlational analyses

r

correlation

dplyr

purrr

tidyverse

purrr -- 将特定模式中的所有变量循环到运行相关分析