R，将组 ID 分配给另一个观察 ID 变量中的值组合

Question

假设我有几年的个人数据，其中个人ID（indiv）可以识别跨年的个人，但家庭ID（househ）只能分配一年内的家庭关系。

示例：

libary(dplyr)
test_01 <- data.frame(indiv=c(1,2,3,4,5,6),househ=c(1,1,2,3,4,4),time=rep(1,6)) #1&2 form a household, 3 and 4 are single, 5&6 form a household
test_02 <- data.frame(indiv=c(2,3,4,5,6,7),househ=c(1,2,2,3,3,4),time=rep(2,6)) #1 exits, so 2 is now a new household, 3&4 now form a new household, 5&6 still do, 7 enters
test_03 <- data.frame(indiv=c(2,3,4,5,7,8,9,10),househ=c(1,2,2,3,4,5,5,6),time=rep(3,8)) #according to logic above
data_test_panel <- bind_rows(test_01,test_02,test_03)

期望的时间一致的家庭变量是：

data_test_panel$true_household <- c(1,1,2,3,4,4,5,6,6,4,4,7,5,6,6,8,7,9,9,10) 到目前为止我尝试过：

library(data.table)
setDT(data_test_panel)[,cons_household := .GRP,.(time,househ)] # where cons_household is the new household ID. However, this doesn't give the same household ID across time but assigns new values for every appearance of a household.

衷心感谢您的帮助！ /塞维林

Answer 1

这应该可以。基本上创建一个家庭中独特个人的列表栏。然后我们可以通过该列获取组号。

dplyr 方法

这好多了：

library(dplyr)
data_test_panel |>
    mutate(
        indiv_in_household = list(indiv),
        .by = c(househ, time)
    ) |>
    mutate(
        cons_household = cur_group_id(),
        .by = indiv_in_household
    )

数据表方法

data_test_panel$indiv_in_household <- data_test_panel[
    , .(split(
        rep(indiv, length(indiv)),
        rep(seq(length(indiv)), each = length(indiv))
    )),
    .(househ, time)
]$V1


head(data_test_panel)
#    indiv househ  time true_household indiv_in_household
#    <num>  <num> <num>          <num>             <list>
# 1:     1      1     1              1                1,2
# 2:     2      1     1              1                1,2
# 3:     3      2     1              2                  3
# 4:     4      3     1              3                  4
# 5:     5      4     1              4                5,6
# 6:     6      4     1              4                5,6

data_test_panel[, cons_household := .GRP, sapply(indiv_in_household, toString)]

data_test_panel[cons_household != true_household]
# Empty data.table (0 rows and 6 cols): indiv,househ,time,true_household,cons_household,indiv_in_household