如何对多个变量的 2 个两个时间点（前、后）运行配对 t.test，而不必列出每个变量？

Question

您好，我是 R 新手，正在尝试运行配对 t.test。

目前已使用pivot_longer从宽格式重塑数据，而不是在单独的列中进行测试，因为我已经有一个时间点列，但不确定如何从这里继续。如何根据时间点和测试运行 t.test？

样本数据集：

   id    timepoint  test          score
  <chr>  <chr>     <chr>           <dbl>
1 101     Pre    testA            35
2 101     Pre    testB            37
3 101     Pre    testC            41
4 101     Post   testA            39
5 101     Post   testB            16
6 101     Post   testC            33
7 102     Pre    testA            27
8 102     Pre    testB            21
9 102     Pre    testC            25

是否也可以一次性完成所有测试？

我使用pivot_longer之前的数据集示例

id  timepoint testA testB testC 
<chr>  <chr>  <dbl> <dbl>  <dbl>       
1 101   Pre    35     37     41
2 101   Post   39     16     33
3 102   Pre    27     21     25

谢谢！

目前我只知道：t.test(pre_test_scores, post_test_scores,paired = TRUE)

编辑：为了更加清晰，包含了一个可重现的示例数据集

set.seed(123)
id <- sample(100:150, 40)

id_rep <- rep(id, each = 6) # 2 timepoints * 3 subscales = 6
timepoints <- rep(c("pre", "post"), each = 3, times = 40)
subscales <- rep(c("emotional_wellbeing", "physical_wellbeing", "overall_wellbeing"), times = 80)


set.seed(456) # Setting seed for reproducibility of scores
scores <- sample(1:100, 240, replace = TRUE) # Assuming scores are between 1 and 100


data <- data.frame(ID = id_rep, Timepoint = timepoints, Subscale = subscales, Score = scores)

print(data)

     ID Timepoint            Subscale Score
1   130       pre emotional_wellbeing    35
2   130       pre  physical_wellbeing    38
3   130       pre   overall_wellbeing    85
4   130      post emotional_wellbeing    27
5   130      post  physical_wellbeing    25
6   130      post   overall_wellbeing    78
7   114       pre emotional_wellbeing    31
8   114       pre  physical_wellbeing    73
9   114       pre   overall_wellbeing    79
10  114      post emotional_wellbeing    90
11  114      post  physical_wellbeing    83
12  114      post   overall_wellbeing    43
13  113       pre emotional_wellbeing    87
14  113       pre  physical_wellbeing     8
15  113       pre   overall_wellbeing    14
16  113      post emotional_wellbeing    13
17  113      post  physical_wellbeing    69
18  113      post   overall_wellbeing    52
19  102       pre emotional_wellbeing    30
20  102       pre  physical_wellbeing    47
21  102       pre   overall_wellbeing    25
22  102      post emotional_wellbeing    38
23  102      post  physical_wellbeing    40
24  102      post   overall_wellbeing    34
25  141       pre emotional_wellbeing    27
26  141       pre  physical_wellbeing    68
27  141       pre   overall_wellbeing    28
28  141      post emotional_wellbeing    10
29  141      post  physical_wellbeing    31
30  141      post   overall_wellbeing    65
31  142       pre emotional_wellbeing    74
32  142       pre  physical_wellbeing    25
33  142       pre   overall_wellbeing    43
34  142      post emotional_wellbeing    20
35  142      post  physical_wellbeing    30
36  142      post   overall_wellbeing    75
37  136       pre emotional_wellbeing    88
38  136       pre  physical_wellbeing    34
39  136       pre   overall_wellbeing    38
40  136      post emotional_wellbeing    94
41  136      post  physical_wellbeing    99
42  136      post   overall_wellbeing    15
43  148       pre emotional_wellbeing    97
44  148       pre  physical_wellbeing     4
45  148       pre   overall_wellbeing    84
46  148      post emotional_wellbeing    16
47  148      post  physical_wellbeing    25
48  148      post   overall_wellbeing    62
49  124       pre emotional_wellbeing    66
50  124       pre  physical_wellbeing    87
51  124       pre   overall_wellbeing    27
52  124      post emotional_wellbeing    80
53  124      post  physical_wellbeing    80
54  124      post   overall_wellbeing    88

Answer 1

如果问题想要在配对的基础上使用单个 p 值来测试 pre = post，则首先创建一个宽数据框，然后运行

t.test

或

lm

，如图所示 - 两者都给出相同的 p 值。

library(tidyr)

wide <- data %>%
  pivot_wider(id_cols = c(ID, Subscale),
     names_from = Timepoint, values_from = Score)

with(wide, t.test(pre, post, paired = TRUE))
# or
summary(lm(post - pre ~ 1, wide))