在ddply中使用ifelse和transform

Question

我试图用ddply与transform填充一个新的变量(summary_Date)与变量数据框ID和Date.根据使用ifelse以下方法评估的工件的长度来选择变量的值:

如果给定月份的ID少于五个,我希望summary_Date通过将日期四舍五入到最接近的月份来计算(使用round_date包裹lubridate); 如果给定月份的ID有超过五个观察值,我希望它summary_Date只是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

该ifelse语句似乎有效,但'summary_Date'中的值似乎是为变换正在处理的子集计算的第一个值,而不是行特定值.例如,在第3行中,summary_Date应该是1962-04-01因为日期1962-03-27 12:00:00'应该向上舍入(因为子集中的行少于5行),而是在该子集的所有行中重复第一个计算值summary_Date(1962-03-01).

编辑:我的灵感来自里卡多的回答,用data.table两步来试试ddply.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")
)