在rstudio中有效地组合多个data.table排列

我有一个data.table，其中包含任意用户ID和测量的因变量：

library(data.table)

data <- data.table(user_id = round(runif(10,0,100),0), dv = round(runif(10, 0, 10),1))
                   

库（data.table）
数据我认为你不太可能发现巨大的速度提升，至少除非你有非常大的数据。但是如果您避免了
cbind
调用，那么您可以编写一个较短且速度不到两倍的版本。我还包括一个稍微快一点的。注意，我将样本大小增加到200，以提供更有意义的基准

library(data.table)
                   
N <- 200
data <- data.table(user_id = round(runif(N,0,100),0), dv = round(runif(N, 0, 10),1))

method1 <- function() {
  for(n in 1:nrow(data)){
    ifelse(n > 1,
           data_combined <- rbindlist(list(data_combined,
                                           cbind(data[1:n, ], N = n))),
           data_combined <- cbind(data[1:n, ], N = n)
    )
  }
  return(data_combined)
}

method2 <- function() {
  data_combined <- lapply(1:nrow(data), function(i) data.table(data[1:i], N=i))
  data_combined <- rbindlist(data_combined)
  return(data_combined)
}

method3 <- function() {
  data_combined <- list()
  for (i in 1:nrow(data)) {
    data_combined[[i]] <- data[1:i]
    data_combined[[i]][, N := i]
  }
  data_combined <- rbindlist(data_combined)
  return(data_combined)
}

基准：

library(microbenchmark)
microbenchmark(method1(), method2(), method3())
#> Unit: milliseconds
#>       expr       min        lq      mean    median        uq      max neval cld
#>  method1() 131.94106 150.40669 166.28043 157.02466 165.40358 408.1812   100   c
#>  method2()  82.57354 100.94865 109.82735 106.36961 109.65404 202.5385   100  b 
#>  method3()  73.79046  90.20843  96.89102  96.14839  99.36719 193.2423   100 a

非常感谢。您的第三种方法正是我正在寻找的那种方法，我还没有完全弄清楚，因为我还在学习data.table包。虽然我也对通过尽可能消除“for”循环来加快迭代感兴趣，但放弃cbind&ifelse调用正是出于这个原因我感兴趣的其他选择。
library(data.table)
                   
N <- 200
data <- data.table(user_id = round(runif(N,0,100),0), dv = round(runif(N, 0, 10),1))

method1 <- function() {
  for(n in 1:nrow(data)){
    ifelse(n > 1,
           data_combined <- rbindlist(list(data_combined,
                                           cbind(data[1:n, ], N = n))),
           data_combined <- cbind(data[1:n, ], N = n)
    )
  }
  return(data_combined)
}

method2 <- function() {
  data_combined <- lapply(1:nrow(data), function(i) data.table(data[1:i], N=i))
  data_combined <- rbindlist(data_combined)
  return(data_combined)
}

method3 <- function() {
  data_combined <- list()
  for (i in 1:nrow(data)) {
    data_combined[[i]] <- data[1:i]
    data_combined[[i]][, N := i]
  }
  data_combined <- rbindlist(data_combined)
  return(data_combined)
}

all(method1() == method2())
#> [1] TRUE
all(method2() == method3())
#> [1] TRUE

library(microbenchmark)
microbenchmark(method1(), method2(), method3())
#> Unit: milliseconds
#>       expr       min        lq      mean    median        uq      max neval cld
#>  method1() 131.94106 150.40669 166.28043 157.02466 165.40358 408.1812   100   c
#>  method2()  82.57354 100.94865 109.82735 106.36961 109.65404 202.5385   100  b 
#>  method3()  73.79046  90.20843  96.89102  96.14839  99.36719 193.2423   100 a