如何对 data.frame 进行子集化?

How to subset a data.frame?

我有这样的数据集

a <- data.frame(var1 = c("patientA", "patientA", "patientA", "patientB", "patientB", "patientB", "patientB"),
                var2 = as.Date(c("2015-01-02","2015-01-04","2015-02-02","2015-02-06","2015-01-02","2015-01-07","2015-04-02")),
                var3 = c(F, T, F, F, F, T, F)               
                )
sequ <- rle(as.character(a$var1))
a$sequ <- sequence(sequ$lengths)

生产

> a
      var1       var2  var3 sequ
1 patientA 2015-01-02 FALSE    1
2 patientA 2015-01-04  TRUE    2
3 patientA 2015-02-02 FALSE    3
4 patientB 2015-02-06 FALSE    1
5 patientB 2015-01-02 FALSE    2
6 patientB 2015-01-07  TRUE    3
7 patientB 2015-04-02 FALSE    4

我如何subset/filter这个数据集,以便我得到所有 var3 == TRUE 和 var2 日期值大于 var3 == TRUE 的行中的行(患者,var1?我试过

subset(a, (var3 == TRUE) & (var2 > var3))

但这不会产生正确的结果集。正确的是

#       var1       var2  var3 sequ
# 1 patientA 2015-01-04  TRUE    2
# 2 patientA 2015-02-02 FALSE    3
# 3 patientB 2015-02-06 FALSE    1
# 4 patientB 2015-01-07  TRUE    3
# 5 patientB 2015-04-02 FALSE    4

您可以尝试 data.table。在这里,我们将 'data.frame' 转换为 'data.table' (setDT(a)),按 'var1' 分组,我们得到大于或等于 'var2' 元素的逻辑索引到相应的 'var2' 个元素,其中 'var3' 为 TRUE 并子集数据集 .SD

library(data.table)
setDT(a)[,.SD[var2 >= var2[var3]], var1]
#       var1       var2  var3 sequ
#1: patientA 2015-01-04  TRUE    2
#2: patientA 2015-02-02 FALSE    3
#3: patientB 2015-02-06 FALSE    1
#4: patientB 2015-01-07  TRUE    3
#5: patientB 2015-04-02 FALSE    4

使用base R的选项(假设数据按'var1'排序)

a[with(a, var2>=rep(var2[var3], table(var1))),]
#      var1       var2  var3 sequ
#2 patientA 2015-01-04  TRUE    2
#3 patientA 2015-02-02 FALSE    3
#4 patientB 2015-02-06 FALSE    1
#6 patientB 2015-01-07  TRUE    3
#7 patientB 2015-04-02 FALSE    4

我在 var3TRUE 时添加了一个包含日期的列,根据它进行过滤,然后将其放在末尾。

library(dplyr)

a %>% group_by(var1)%>%
    mutate(truedate = first(var2[var3])) %>%
    filter(var2 >= truedate) %>%
    select(-truedate)

# Source: local data frame [5 x 4]
# Groups: var1

#       var1       var2  var3 sequ
# 1 patientA 2015-01-04  TRUE    2
# 2 patientA 2015-02-02 FALSE    3
# 3 patientB 2015-02-06 FALSE    1
# 4 patientB 2015-01-07  TRUE    3
# 5 patientB 2015-04-02 FALSE    4

A base-R 解决方案:首先,不要理会你的rle/sequ 事情。相反,对数据进行排序:

a <- a[order(a$var1,a$var2),]

查找所选行:

myrows <- tapply(
  1:nrow(a),
  a$var1,
  function(ivec){
    istar <- ivec[a$var3[ivec]]
    ivec[ivec>=istar]
  })

子集 a[unlist(myrows),]