在 Tidy Data 中采用一对多的组间差异

Question

我有一个整洁的数据集，类似于 Introducing tidyr blog post 中的听力示例，但我在药物下还有一个额外的 "placebo" 组，我可以像这样构建

library(dplyr)
library(tidyr)
messy <- data.frame(
  name = c("Wilbur", "Petunia", "Gregory"),
  a = c(67, 80, 64),
  b = c(56, 90, 50),
  p = c(60, 70, 60)    # this is the new 'placebo' drug
)
tidy <- messy %>% 
        gather(drug, heartrate, a:p)

假设我从整洁的数据开始，我的目标是创建一个名为 "diff.p" 的新变量，它是每种药物和安慰剂的观察值之间的差异。结果应如下所示

tidy$diff.p <- c(7,10,4,-4,20,-10,0,0,0)
tidy

似乎 ave and/or mutate 可能是解决问题的好途径（或者构建一个新的数据框？），但我需要一些关于最佳实践的额外指导.

Answer 1

看来你可以很容易地用第二个 tidy:

tidy2 <- messy %>%
  mutate(a = a-p, b = b-p, p = 0) %>%
  gather(drug, diff.p, a:p)

left_join(tidy, tidy2, by = c("name", "drug"))
#      name drug heartrate diff.p
# 1  Wilbur    a        67      7
# 2 Petunia    a        80     10
# 3 Gregory    a        64      4
# 4  Wilbur    b        56     -4
# 5 Petunia    b        90     20
# 6 Gregory    b        50    -10
# 7  Wilbur    p        60      0
# 8 Petunia    p        70      0
# 9 Gregory    p        60      0

Answer 2

在 dplyr 链中，您可以按 name 分组，然后从 heartrate 中减去 heartrate[drug=="p"]：

tidy = tidy %>% group_by(name) %>% 
  mutate(diff.p2 = heartrate - heartrate[drug=="p"])

     name  drug heartrate diff.p diff.p2
   <fctr> <chr>     <dbl>  <dbl>   <dbl>
1  Wilbur     a        67      7       7
2 Petunia     a        80     10      10
3 Gregory     a        64      4       4
4  Wilbur     b        56     -4      -4
5 Petunia     b        90     20      20
6 Gregory     b        50    -10     -10
7  Wilbur     p        60      0       0
8 Petunia     p        70      0       0
9 Gregory     p        60      0       0

Answer 3

另一种选择是data.table

library(data.table)
melt(setDT(messy), id.var = "name", variable.name = "drug", 
 value.name = "heartrate")[, diff.p2 := heartrate - heartrate[drug=="p"]][]
#      name drug heartrate diff.p2
#1:  Wilbur    a        67       7
#2: Petunia    a        80      10
#3: Gregory    a        64       4
#4:  Wilbur    b        56      -4
#5: Petunia    b        90      20
#6: Gregory    b        50     -10
#7:  Wilbur    p        60       0
#8: Petunia    p        70       0
#9: Gregory    p        60       0

在 Tidy Data 中采用一对多的组间差异

Taking one to many difference Between Groups in Tidy Data

r

tidyr

tidyverse