r - 在 ddply 中使用 ifelse 和转换

Question

我正在尝试使用 ddply与 transform使用变量 summary_Date 在数据框中填充新变量 ( ID )和 Date .变量的值是根据使用 ifelse 评估的片段的长度选择的。 :

如果给定月份中某个 ID 的观察值少于五个，我想要 summary_Date通过将日期四舍五入到最近的月份来计算(使用 round_date 来自包 lubridate )；如果给定月份中某个 ID 的观察值超过五个，我想要 summary_Date简单地成为 Date .

require(plyr)
require(lubridate)

test.df <- structure(
  list(ID = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,1, 1, 1, 1, 1, 1, 1, 1, 1
                , 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,2, 2, 2, 2, 2, 2, 2, 2)
       , Date = structure(c(-247320000, -246196800, -245073600, -243864000
                            , -242654400, -241444800, -126273600, -123595200
                            , -121176000, -118497600, 1359385200, 1359388800
                            , 1359392400, 1359396000, 1359399600, 1359403200
                            , 1359406800, 1359410400, 1359414000, 1359417600
                            , 55598400, 56116800, 58881600, 62078400, 64756800
                            , 67348800, 69854400, 72964800, 76161600, 79012800
                            , 1358589600, 1358676000, 1358762400, 1358848800
                            , 1358935200, 1359021600, 1359108000, 1359194400
                            , 1359280800, 1359367200), tzone = "GMT"
                          , class = c("POSIXct", "POSIXt"))
       , Val=rnorm(40))
  , .Names = c("ID", "Date", "Val"), row.names = c(NA, 40L)
  , class = "data.frame")

test.df <- ddply(test.df, .(ID, floor_date(Date, "month")), transform
                 , summary_Date=as.POSIXct(ifelse(length(ID)<5
                                                  , round_date(Date, "month")
                                                  ,Date)
                                           , origin="1970-01-01 00:00.00"
                                           , tz="GMT")
                 # Included length_x to easily see the length of the subset
                 , length_x = length(ID))

head(test.df,5)
#   floor_date(Date, "month") ID                Date        Val summary_Date length_x
# 1                1962-03-01  1 1962-03-01 12:00:00 -0.1037988   1962-03-01        3
# 2                1962-03-01  1 1962-03-14 12:00:00  0.2923056   1962-03-01        3
# 3                1962-03-01  1 1962-03-27 12:00:00  0.4435410   1962-03-01        3
# 4                1962-04-01  1 1962-04-10 12:00:00  0.1159164   1962-04-01        2
# 5                1962-04-01  1 1962-04-24 12:00:00  2.9824075   1962-04-01        2

ifelse语句似乎有效，但 'summary_Date' 中的值似乎是为转换正在处理的子集计算的第一个值，而不是特定于行的值。例如在第 3 行， summary_Date应该是 1962-04-01因为日期 1962-03-27 12:00:00'应该四舍五入(因为子集中少于五行)，而是取第一个计算值 summary_Date ( 1962-03-01 ) 在该子集中的所有行中重复。

编辑:我的灵感来自 Ricardo 使用 data.table 的回答用 ddply 分两步试试.它也有效:

test.df <- ddply(test.df, .(ID, floor_date(Date, "month")), transform
                 , length_x = length(ID))

test.df <- ddply(test.df, .(ID, floor_date(Date, "month")), transform
                 , summary_Date=as.POSIXct(ifelse(length_x<5
                                                  , round_date(Date, "month")
                                                  ,Date)
                                           , origin="1970-01-01 00:00.00"
                                           , tz="GMT"))

head(test.df,5)[c(1,3:7)]
#   floor_date(Date, "month") ID                Date        Val length_x summary_Date
# 1                1962-03-01  1 1962-03-01 12:00:00 -0.1711212        3   1962-03-01
# 2                1962-03-01  1 1962-03-14 12:00:00 -0.1531571        3   1962-03-01
# 3                1962-03-01  1 1962-03-27 12:00:00  0.1256238        3   1962-04-01
# 4                1962-04-01  1 1962-04-10 12:00:00  1.4481225        2   1962-04-01
# 5                1962-04-01  1 1962-04-24 12:00:00 -0.6508731        2   1962-05-01

Answer 1

一步ddply解决方案(也作为评论发布)

ddply(test.df, .(ID, floor_date(Date, "month")), mutate, 
  length_x = length(ID), 
  summary_Date=as.POSIXct(ifelse(length_x < 5, round_date(Date, "month") ,Date)
    , origin="1970-01-01 00:00.00", tz="GMT")
)