在 data.table 中对数值和因子变量进行分组

Question

我正在尝试对 data.table 的多个数字和因子列进行分组。 这是一些数据：

library(data.table)
dt <- data.table(Zeit = c(117.9, 118, 118, 118, 118.1, 118.1, 118.2, 118.2, 118.2, 118.3, 118.3), 
                 a = factor(c(15, 15, 1, 0, 0, 0, 1, 1, 15, 15, 15)), 
                 b = factor(c(1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1)))

我正在使用这个功能：

compr_uniq_data <- function(ecuData) {
  ecuData <- ecuData[, lapply(.SD, function(x) if (is.numeric(x)) mean(x, na.rm = TRUE)
                   else unlist(unique(x[!is.na(x)]))), by = Zeit]
  return(ecuData)
}

compr_uniq_data(dt)

但我收到以下错误： Supplied 2 items for column 2 of group 2 which has 3 rows. The RHS length must either be 1 (single values are ok) or match the LHS length exactly. If you wish to 'recycle' the RHS please use rep() explicitly to make this intent clear to readers of your code.

我做错了什么？ 如果我只使用一个因子列，它就可以工作。

预期的结果应该是这样的：

dtres <- data.table(Zeit = c(117.9, 118, 118, 118, 118.1, 118.2, 118.2, 118.3), 
                    a = c(15, 15, 1, 0, 0, 1, 15, 15), 
                    b = c(1, 1, 1, 0, 0, 1, 1,  1))

Answer 1

length 应该与 data.table/data.frame 列相同，因此请将其包装在 list 或 paste 中

compr_uniq_data <- function(ecuData) {
  ecuData <- ecuData[, lapply(.SD, function(x)
       if (is.numeric(x)) mean(x, na.rm = TRUE)
                else list(unique(x[!is.na(x)]))), by = Zeit]
    return(ecuData)
 }

compr_uniq_data(dt)
#   Zeit        a    b
#1: 117.9       15    1
#2: 118.0 15, 1, 0 1, 0
#3: 118.1        0    0
#4: 118.2    1, 15    1
#5: 118.3       15    1

更新

基于 cmets，我们可以更改函数以将不是numeric 的数据集列的unique 作为一个块，然后将on 与numeric 列（如果存在）按“Zeit”分组'

compr_uniq_data <- function(ecuData) {
     i1 <-  ecuData[, unlist(lapply(.SD, is.numeric))]
    
     nm1 <- setdiff(names(ecuData)[!i1], "Zeit")
     
     out1 <- unique(ecuData, by = c(nm1, "Zeit"))
     
     if(any(i1[names(i1) != "Zeit"])) {
          nm2 <- setdiff(names(ecuData)[i1], "Zeit")
          out2 <- ecuData[, lapply(.SD, mean, na.rm = TRUE), 
                 by = Zeit, .SDcols = nm2]
          out1 <- out1[out2, on = .(Zeit)]
     } else {
         out1
     }
    out1

}

-测试

compr_uniq_data(dt)
#    a b  Zeit
#1: 15 1 117.9
#2: 15 1 118.0
#3:  1 1 118.0
#4:  0 0 118.0
#5:  0 0 118.1
#6:  1 1 118.2
#7: 15 1 118.2
#8: 15 1 118.3

---

或者另一种选择是将replace duplicated 元素与NA 然后删除那些包含所有NAs 的行用于除“Zeit”之外的列

compr_uniq_data <- function(ecuData) {
  ecuData <- ecuData[, lapply(.SD, function(x)
       if (is.numeric(x)) mean(x, na.rm = TRUE)
                else replace(x, duplicated(x), NA)), by = Zeit]
  
  ecuData[!ecuData[, Reduce(`&`, lapply(.SD, is.na)),
      .SDcols = setdiff(names(ecuData), 'Zeit')]]
 }


compr_uniq_data(dt)