有没有办法“合并”两列，其中新列的值是具有特定值的原始列的名称，分组明智？
Posted
技术标签:
【中文标题】有没有办法“合并”两列，其中新列的值是具有特定值的原始列的名称，分组明智？【英文标题】：Is there a way to 'merge' two columns, where the values of new column are the name of the original column that had a specific value, group wise? 【发布时间】：2020-03-26 01:26:05 【问题描述】：我有一个数据框（将其称为“df”），其中包含相当数量的变量（数字、逻辑和字符），代表不同细胞类型从特定介质移动到另一种介质的实验，以及在特定时间对细胞进行量化。第一列和第二列分别包含“源”介质的名称和细胞移动到的介质的名称；第三列描述了活动被量化的时间，第四列是细胞类型，第五列是测量的活动，这就是有趣的地方。
我有两个主要问题，第一个是要知道是否有“R-esque”方式来做我所做的以获得第六列，其中包含值的增加/减少（百分比） 'Activity' 相对于上一行中的活动，但以组的方式（每个组由 Cell.Type、Pre.Medium 和 Time 的组合组成），这就是为什么每次 Time 的值为零时它的值都是 NA .
假设这是我的数据框（我已对其进行了简化以使我的问题更清楚）：
df <- structure(list(Pre.Medium = c("Medium1", "Medium1", "Medium1",
"Medium2", "Medium2", "Medium2", "Medium1", "Medium1", "Medium1",
"Medium2", "Medium2", "Medium2"), Pos.Medium = c("Medium2", "Medium2",
"Medium2", "Medium1", "Medium1", "Medium1", "Medium2", "Medium2",
"Medium2", "Medium1", "Medium1", "Medium1"), Time = c(0, 2, 4,
0, 2, 4, 0, 2, 4, 0, 2, 4), Cell.Type = c("Cell_A", "Cell_A",
"Cell_A", "Cell_A", "Cell_A", "Cell_A", "Cell_B", "Cell_B", "Cell_B",
"Cell_B", "Cell_B", "Cell_B"), Activity = c(0.5, 1, 2, 2, 1,
0.5, 0.2, 0.8, 0.2, 0.2, 0.2, 0.4), Percent.Increase = c(NA,
100, 100, NA, -50, -50, NA, 300, -75, NA, 0, 100), Primary.Increase = c(NA,
TRUE, FALSE, NA, TRUE, FALSE, NA, TRUE, FALSE, NA, FALSE, FALSE
), Secondary.Increase = c(NA, FALSE, FALSE, NA, FALSE, FALSE,
NA, FALSE, FALSE, NA, FALSE, TRUE)), class = c("spec_tbl_df",
"tbl_df", "tbl", "data.frame"), row.names = c(NA, -12L), problems = structure(list(
row = 1L, col = NA_character_, expected = "8 columns", actual = "9 columns",
file = "'new 2'"), row.names = c(NA, -1L), class = c("tbl_df",
"tbl", "data.frame")), spec = structure(list(cols = list(Pre.Medium = structure(list(), class = c("collector_character",
"collector")), Pos.Medium = structure(list(), class = c("collector_character",
"collector")), Time = structure(list(), class = c("collector_double",
"collector")), Cell.Type = structure(list(), class = c("collector_character",
"collector")), Activity = structure(list(), class = c("collector_double",
"collector")), Percent.Increase = structure(list(), class = c("collector_double",
"collector")), Primary.Increase = structure(list(), class = c("collector_logical",
"collector")), Secondary.Increase = structure(list(), class = c("collector_logical",
"collector"))), default = structure(list(), class = c("collector_guess",
"collector")), skip = 1), class = "col_spec"))
### Pre.Med Pos.Med Time Cell.Type Activity Percent.Increase Primary.Increase Secondary.Increase
### Medium1 Medium2 0 Cell_A 0.5 NA NA NA
### Medium1 Medium2 2 Cell_A 1 100 TRUE FALSE
### Medium1 Medium2 4 Cell_A 2 100 FALSE FALSE
### Medium2 Medium1 0 Cell_A 2 NA NA NA
### Medium2 Medium1 2 Cell_A 1 -50 TRUE FALSE
### Medium2 Medium1 4 Cell_A 0.5 -50 FALSE FALSE
### Medium1 Medium2 0 Cell_B 0.2 NA NA NA
### Medium1 Medium2 2 Cell_B 0.8 300 TRUE FALSE
### Medium1 Medium2 4 Cell_B 0.2 -75 FALSE FALSE
### Medium2 Medium1 0 Cell_B 0.2 NA NA NA
### Medium2 Medium1 2 Cell_B 0.2 0 FALSE FALSE
### Medium2 Medium1 4 Cell_B 0.4 100 FALSE TRUE
我通过使用 group_by 和 mutate 函数，然后使用 lag 函数来计算上一行和上一行的增加/减少，有没有更好的方法呢？对于我的具体情况，滞后就足够了，但是如果我在每个“组”中有超过三个时间测量并且需要远远落后于计算呢？使用我的方法，在某些时候我不得不使用诸如 lag(lag(lag(lag(lag((Activity / lag(Activity)) - 1) * 100)))) 之类的东西。
另一件事是我无法以任何方式弄清楚的事情，它是将我的“宽”数据集变成一个长数据集，方法是将我的列“Primary.Increase”和“Secondary.Increase”到名为“Increase.Type”的列中，对于每个组（Cell.Type、Pre.Med 和 Time 的组合），其值将包含在列名（Primary.Response 或 Secondary.Response）中，其中其成员之一的值为 TRUE。它应该看起来像这样：
df <- structure(list(Pre.Med = c("Medium1", "Medium1", "Medium1", "Medium2",
"Medium2", "Medium2", "Medium1", "Medium1", "Medium1", "Medium2",
"Medium2", "Medium2"), Pos.Med = c("Medium2", "Medium2", "Medium2",
"Medium1", "Medium1", "Medium1", "Medium2", "Medium2", "Medium2",
"Medium1", "Medium1", "Medium1"), Time = c(0, 2, 4, 0, 2, 4,
0, 2, 4, 0, 2, 4), Cell.Type = c("Cell_A", "Cell_A", "Cell_A",
"Cell_A", "Cell_A", "Cell_A", "Cell_B", "Cell_B", "Cell_B", "Cell_B",
"Cell_B", "Cell_B"), Activity = c(0.5, 1, 2, 2, 1, 0.5, 0.2,
0.8, 0.2, 0.2, 0.2, 0.4), Percent.Inc = c(NA, 100, 100, NA, -50,
-50, NA, 300, -75, NA, 0, 100), Increase.Type = c("Primary.Increase",
"Primary.Increase", "Primary.Increase", "Primary.Increase", "Primary.Increase",
"Primary.Increase", "Primary.Increase", "Primary.Increase", "Primary.Increase",
"Secondary.Increase", "Secondary.Increase", "Secondary.Increase"
)), class = c("spec_tbl_df", "tbl_df", "tbl", "data.frame"), row.names = c(NA,
-12L), spec = structure(list(cols = list(Pre.Med = structure(list(), class = c("collector_character",
"collector")), Pos.Med = structure(list(), class = c("collector_character",
"collector")), Time = structure(list(), class = c("collector_double",
"collector")), Cell.Type = structure(list(), class = c("collector_character",
"collector")), Activity = structure(list(), class = c("collector_double",
"collector")), Percent.Inc = structure(list(), class = c("collector_double",
"collector")), Increase.Type = structure(list(), class = c("collector_character",
"collector"))), default = structure(list(), class = c("collector_guess",
"collector")), skip = 1), class = "col_spec"))
### Pre.Med Pos.Med Time Cell.Type Activity Percent.Inc Increase.Type
### Medium1 Medium2 0 Cell_A 0.5 NA Primary.Increase
### Medium1 Medium2 2 Cell_A 1 100 Primary.Increase
### Medium1 Medium2 4 Cell_A 2 100 Primary.Increase
### Medium2 Medium1 0 Cell_A 2 NA Primary.Increase
### Medium2 Medium1 2 Cell_A 1 -50 Primary.Increase
### Medium2 Medium1 4 Cell_A 0.5 -50 Primary.Increase
### Medium1 Medium2 0 Cell_B 0.2 NA Primary.Increase
### Medium1 Medium2 2 Cell_B 0.8 300 Primary.Increase
### Medium1 Medium2 4 Cell_B 0.2 -75 Primary.Increase
### Medium2 Medium1 0 Cell_B 0.2 NA Secondary.Increase
### Medium2 Medium1 2 Cell_B 0.2 0 Secondary.Increase
### Medium2 Medium1 4 Cell_B 0.4 100 Secondary.Increase
首先有没有办法做到这一点？我假设是这样，但到目前为止我还不能做到：/ 我是一名生物学本科生，对 R 比较陌生，我很喜欢你可以用它做什么，但我离擅长它还有很长的路要走。
非常感谢任何帮助。
【问题讨论】：
【参考方案1】：我不确定我是否理解第一个问题。 如果您执行以下操作：
library(dplyr)
df %>%
group_by(Cell.Type, Pre.Medium, Pos.Medium) %>%
arrange(Time, .by_group = TRUE) %>% # remove if Time is always ascending
mutate(Percent.Increase = ((Activity / lag(Activity)) - 1) * 100)
Percent.Increase的计算是向量化的，
所以Activity 有多长并不重要
（另请参阅下面我的最后解释）。
对于第二个问题， 如果我理解正确， 你可以这样做：
df %>%
group_by(Cell.Type, Pre.Medium, Pos.Medium) %>%
mutate(Increase.Type = if (any(Secondary.Increase, na.rm = TRUE)) "Secondary.Increase" else "Primary.Increase") %>%
select(-(Primary.Increase:Secondary.Increase))
# A tibble: 12 x 7
# Groups: Cell.Type, Pre.Medium, Pos.Medium [4]
Pre.Medium Pos.Medium Time Cell.Type Activity Percent.Increase Increase.Type
<chr> <chr> <dbl> <chr> <dbl> <dbl> <chr>
1 Medium1 Medium2 0 Cell_A 0.5 NA Primary.Increase
2 Medium1 Medium2 2 Cell_A 1 100 Primary.Increase
3 Medium1 Medium2 4 Cell_A 2 100 Primary.Increase
4 Medium2 Medium1 0 Cell_A 2 NA Primary.Increase
5 Medium2 Medium1 2 Cell_A 1 -50 Primary.Increase
6 Medium2 Medium1 4 Cell_A 0.5 -50 Primary.Increase
7 Medium1 Medium2 0 Cell_B 0.2 NA Primary.Increase
8 Medium1 Medium2 2 Cell_B 0.8 300 Primary.Increase
9 Medium1 Medium2 4 Cell_B 0.2 -75 Primary.Increase
10 Medium2 Medium1 0 Cell_B 0.2 NA Secondary.Increase
11 Medium2 Medium1 2 Cell_B 0.2 0 Secondary.Increase
12 Medium2 Medium1 4 Cell_B 0.4 100 Secondary.Increase
mutate 内部的转换会看到组中的所有值，
所以any(Secondary.Increase, na.rm = TRUE) 一次接收所有元素，
如果我们只返回 1 个值，
它将被复制以适应组大小。
【讨论】：
以上是关于有没有办法“合并”两列，其中新列的值是具有特定值的原始列的名称，分组明智？的主要内容，如果未能解决你的问题，请参考以下文章