我有以下组织数据:
EmployeeID <- c(10:15)
Job.Title <- c("Program Manager", "Development Manager", "Developer" , "Developer", "Developer", "Summer Intern")
Level.1 <- c(1,1,1,1,1,1)
Level.2 <- c(2,2,2,2,2,2)
Level.3 <- c("",10,10,10,10,10)
Level.4 <- c("","",11,11,11,11)
Level.5 <- c("","","","","",12)
Level.6 <- c("","","","","","")
Pay.Type <- c("Salary", "Salary", "Salary", "Salary", "Salary", "Hourly")
acme = data.frame(EmployeeID, Job.Title, Level.1, Level.2, Level.3, Level.4, Level.5, Level.6, Pay.Type)
acme
EmployeeID Job.Title Level.1 Level.2 Level.3 Level.4 Level.5 Level.6 Pay.Type
1 10 Program Manager 1 2 Salary
2 11 Development Manager 1 2 10 Salary
3 12 Developer 1 2 10 11 Salary
4 13 Developer 1 2 10 11 Salary
5 14 Developer 1 2 10 11 Salary
6 15 Summer Intern 1 2 10 11 12 Hourly
对于每一行,我需要识别Level.1到Level.6的第一个非NULL值,从Level.6开始,然后是Level.5,然后是Level.4,依此类推。我还需要在同一模式中识别第二个非Null值。每行的标识值需要放在新列中,因此最终表格如下所示:
EmployeeID Job.Title Level.1 Level.2 Level.3 Level.4 Level.5 Level.6 Pay.Type Supervisor Manager
1 10 Program Manager 1 2 Salary 2 1
2 11 Development Manager 1 2 10 Salary 10 2
3 12 Developer 1 2 10 11 Salary 11 10
4 13 Developer 1 2 10 11 Salary 11 10
5 14 Developer 1 2 10 11 Salary 11 10
6 15 Summer Intern 1 2 10 11 12 Hourly 12 11
答案 0 :(得分:4)
我们可以逐行使用apply
并获取所有非空的索引,并选择第一个和第二个值分别获得两列。
acme[, c("Supervisor", "Manager")] <- t(apply(acme[, 8:3], 1,
function(x) c(x[which(x != "")[1]], x[which(x != "")[2]])))
acme
# EmployeeID Job.Title Level.1 Level.2 Level.3 Level.4 Level.5 Level.6 Pay.Type Supervisor Manager
#1 10 Program Manager 1 2 Salary 2 1
#2 11 Development Manager 1 2 10 Salary 10 2
#3 12 Developer 1 2 10 11 Salary 11 10
#4 13 Developer 1 2 10 11 Salary 11 10
#5 14 Developer 1 2 10 11 Salary 11 10
#6 15 Summer Intern 1 2 10 11 12 Hourly 12 11
修改强>
如果有很多列,我们需要找到开始和结束列的索引。我们可以将grep
用于相同的
mincol <- min(grep("Level", colnames(acme)))
maxcol <- max(grep("Level", colnames(acme)))
acme[, c("Supervisor", "Manager")] <- t(apply(acme[, maxcol:mincol], 1,
function(x) c(x[which(x != "")[1]], x[which(x != "")[2]])))
应该有效。
如果我们只需要Supervisor
,我们可以忽略第二部分。
acme[, "Supervisor"] <- t(apply(acme[, maxcol:mincol], 1,
function(x) x[which(x != "")[1]]))
答案 1 :(得分:3)
这是data.table
“单行”:
library(data.table)
setDT(acme)[melt(acme, measure.vars = patterns("Level.\\d"))[value != ""][
order(variable), .(Supervisor = value[.N], Manager = value[.N - 1]), by = EmployeeID],
on = "EmployeeID"][]
EmployeeID Job.Title Level.1 Level.2 Level.3 Level.4 Level.5 Level.6 Pay.Type Supervisor
#1: 10 Program Manager 1 2 Salary 2
#2: 11 Development Manager 1 2 10 Salary 10
#3: 12 Developer 1 2 10 11 Salary 11
#4: 13 Developer 1 2 10 11 Salary 11
#5: 14 Developer 1 2 10 11 Salary 11
#6: 15 Summer Intern 1 2 10 11 12 Hourly 12
Manager
#1: 1
#2: 2
#3: 10
#4: 10
#5: 10
#6: 11
工作原理
data.frame
被强制为data.table
""
的所有行。Level.1
,Level.2
等。)acme
以附加新列 注意: melt()
会发出一条警告消息,指出并非所有级别列都具有相同的数据类型。这是由于在""
data.frame的定义中将整数值与字符(acme
)混合而引起的。最好使用NA
代替""
。顺便说一句:在这种情况下,可以使用na.rm = FALSE
melt()
注意:步骤4中简单的alaphybetical排序最多可以使用9个级别(Level.1
到Level.9
)。如果有更多级别,则必须提取级别编号并强制转换为整数。
答案 2 :(得分:3)
dplyr
和tidyr
依赖于数据重塑的解决方案。
library(tidyverse)
acme %>%
gather('level', 'value', starts_with('Level.')) %>%
group_by(EmployeeID) %>%
filter(value != '') %>%
summarise(Supervisor = last(value),
Manager = nth(value, -2)) %>%
left_join(acme)
答案 3 :(得分:2)
我们可以使用max.col
执行此操作。找到&#39; Level&#39;的索引。列(&#39; i1&#39;),转换&#39; acme&#39;的子集基于&#39; i1&#39;到matrix
(!=""
),应用max.col
并获取last
TRUE值的列索引,减去1以获得倒数第二个TRUE值(&#39; i3&#39;),使用行/列索引提取元素并创建&#39; Supervisor&#39;和经理&#39;列
i1 <- grep("Level\\.\\d+", names(acme))
i2 <- max.col(acme[i1]!="", "last")
i3 <- i2-1
acme$Supervisor <- acme[i1][cbind(1:nrow(acme), i2)]
acme$Manager <- acme[i1][cbind(1:nrow(acme), i3)]
acme
# EmployeeID Job.Title Level.1 Level.2 Level.3 Level.4 Level.5 Level.6 Pay.Type Supervisor Manager
#1 10 Program Manager 1 2 Salary 2 1
#2 11 Development Manager 1 2 10 Salary 10 2
#3 12 Developer 1 2 10 11 Salary 11 10
#4 13 Developer 1 2 10 11 Salary 11 10
#5 14 Developer 1 2 10 11 Salary 11 10
#6 15 Summer Intern 1 2 10 11 12 Hourly 12 11
注意:此解决方案非常简单有效,无需任何不必要的重塑