R 将数据从一列转换为按组累计计数的单个虚拟变量

可以基于值event创建列，如果每行中的列名等于event值，则为其分配1.然后，您可以使用cumsum乘以person.id来累加值.

library(data.table)

categories <- sort(as.character(unique(current$event)))

current[, (categories) := lapply(
                            lapply(categories, function(x) +(event == x)), 
                                 cumsum), .(person.id)][]

#>    person.id     event       date categoryA categoryB categoryC categoryD
#> 1:         1 categoryA 2020-01-01         1         0         0         0
#> 2:         2 categoryA 2020-03-23         1         0         0         0
#> 3:         3 categoryD 2020-09-09         0         0         0         1
#> 4:         1 categoryA 2020-12-30         2         0         0         0
#> 5:         2 categoryC 2021-06-03         1         0         1         0
#> 6:         3 categoryB 2022-03-22         0         1         0         1

如果您更喜欢dcast，并且希望保留顺序并避免日期冲突，请try 在dcast之前添加行ID号.

result  <- dcast(current[, rid := .I], 
                    rid + person.id + date + event ~ event, 
                    fun = length)[, rid := NULL][]

cols <- setdiff(names(result), names(current))

result[, (cols) := lapply(.SD, cumsum), by = person.id, .SDcols = cols][]

#>    person.id       date     event categoryA categoryB categoryC categoryD
#> 1:         1 2020-01-01 categoryA         1         0         0         0
#> 2:         2 2020-03-23 categoryA         1         0         0         0
#> 3:         3 2020-09-09 categoryD         0         0         0         1
#> 4:         1 2020-12-30 categoryA         2         0         0         0
#> 5:         2 2021-06-03 categoryC         1         0         1         0
#> 6:         3 2022-03-22 categoryB         0         1         0         1

不过，我觉得很难相信dcast更快:

library(microbenchmark)

microbenchmark(
  M_categories = {
    categories <- sort(as.character(unique(current$event)))
    current[, (categories) := lapply(
                                lapply(categories, function(x) event == x),
                                cumsum), .(person.id)]
  },
  
  M_dcast = {
    result  <- dcast(current[, rid := .I], 
                        rid + person.id + date + event ~ event, 
                        fun = length)[, rid := NULL][]
    cols <- setdiff(names(result), names(current))
    result[, (cols) := lapply(.SD, cumsum), by = person.id, .SDcols = cols][]
  },
  
  GT_dcats = {
    result = dcast(current, formula = person.id + date ~ event, fun.aggregate = length)
    cols = names(result)[names(result) %like% "category"]
    result[, (cols) := lapply(.SD, cumsum), by = person.id, .SDcols = cols]
  }
  )

Unit: microseconds
         expr    min      lq     mean  median      uq     max neval
 M_categories  563.4  600.45  655.900  626.15  651.95  1302.0   100
      M_dcast 3963.3 4235.40 4627.580 4443.90 4620.25 15476.1   100
     GT_dcats 3355.5 3554.40 4030.594 3711.10 3929.75 14002.1   100

对于较大的数据集，dcast的速度明显较慢:

large_data <- current[sample(nrow(current), 10000, replace = TRUE), ]

microbenchmark(
  M_categories = {
    categories <- sort(as.character(unique(large_data$event)))
    large_data[, (categories) := lapply(
                                lapply(categories, function(x) event == x),
                                cumsum), .(person.id)]
  },

  M_dcast = {
    result  <- dcast(large_data[, rid := .I], 
                        rid + person.id + date + event ~ event, 
                        fun = length)[, rid := NULL][]
    cols <- setdiff(names(result), names(large_data))
    result[, (cols) := lapply(.SD, cumsum), by = person.id, .SDcols = cols][]
  },

  GT_dcats = {
    result = dcast(large_data, formula = person.id + date ~ event, fun.aggregate = length)
    cols = names(result)[names(result) %like% "category"]
    result[, (cols) := lapply(.SD, cumsum), by = person.id, .SDcols = cols]
  }
  )

Unit: milliseconds
         expr     min       lq     mean   median       uq     max neval
 M_categories  1.4199  1.65105  2.23003  1.98575  2.45670 10.9188   100
      M_dcast 26.1525 27.53905 28.46625 28.07925 28.64415 36.4706   100
     GT_dcats  4.3829  4.94760  5.49732  5.44890  5.78995 15.0705   100

^{创建于2024-01-29年第reprex v2.0.2页}