如何按天聚合 POSIXlt 以找到每日峰值？

Question

我有这个问题的变体：Count values in a data set that exceed a threshold in R:

我以几乎随机的时间间隔进行一些温度测量。我想找出（在某个时间范围内）超过特定阈值的天数（多少天）。显然，如果不进行汇总，我可以在同一天获得多次点击（如果多次超过阈值）。但是我不想那样。

原始数据框的一个简短示例如下所示：

                 Time Temp Humidity Notes
1 2015-05-18 16:00:00 26.5       NA  <NA>
2 2015-06-01 15:00:00 26.5       NA  <NA>
3 2015-06-02 16:00:00 28.0       NA  <NA>
4 2015-06-03 16:00:00 28.0       NA  <NA>
5 2015-06-03 17:00:00 30.0       60  <NA>
6 2015-06-05 07:00:00 23.0       NA  <NA>

所以我计算了一个Day 变量（POSIXlt）：

1 2015-05-18 16:00:00 26.5       NA  <NA> 2015-05-18
2 2015-06-01 15:00:00 26.5       NA  <NA> 2015-06-01
3 2015-06-02 16:00:00 28.0       NA  <NA> 2015-06-02
4 2015-06-03 16:00:00 28.0       NA  <NA> 2015-06-03
5 2015-06-03 17:00:00 30.0       60  <NA> 2015-06-03
6 2015-06-05 07:00:00 23.0       NA  <NA> 2015-06-05

我几乎绝望地尝试按天汇总（我没有显示我尝试过的所有变体）：

with(t, aggregate(Temp ~ Day, data=t, FUN=max))
Error in model.frame.default(formula = Temp ~ Day, data = t) : 
  invalid type (list) for variable 'Day'

只有当我将 POSIXlt 显式转换为 POSIXct 时，它才有效（为什么有一个类 POSTXlt 被聚合视为一个列表？）：

> with(t, aggregate(Temp ~ as.POSIXct(Day), data=t, FUN=max))
  as.POSIXct(Day) Temp
1      2015-05-18 26.5
2      2015-06-01 26.5
3      2015-06-02 28.0
4      2015-06-03 30.0
5      2015-06-05 23.0

不幸的是，我在聚合过程中丢失了其他列。我该如何保存它们？

我也不明白这个：

> tt <-with(t, aggregate(Temp ~ as.POSIXct(Day), data=t, FUN=max))
> tt
  as.POSIXct(Day) Temp
1      2015-05-18 26.5
2      2015-06-01 26.5
3      2015-06-02 28.0
4      2015-06-03 30.0
5      2015-06-05 23.0
> str(tt)
'data.frame':   5 obs. of  2 variables:
 $ as.POSIXct(Day): POSIXct, format: "2015-05-18" "2015-06-01" ...
 $ Temp           : num  26.5 26.5 28 30 23
> tt$Temp > 25
[1]  TRUE  TRUE  TRUE  TRUE FALSE
> tt[tt$Temp > 25]
Error in `[.data.frame`(tt, tt$Temp > 25) : undefined columns selected
> tt[tt$Temp > 25,]
  as.POSIXct(Day) Temp
1      2015-05-18 26.5
2      2015-06-01 26.5
3      2015-06-02 28.0
4      2015-06-03 30.0
> t$Temp > 25
[1]  TRUE  TRUE  TRUE  TRUE  TRUE FALSE
> t[t$Temp > 25]
                 Time Temp Humidity Notes        Day
1 2015-05-18 16:00:00 26.5       NA  <NA> 2015-05-18
2 2015-06-01 15:00:00 26.5       NA  <NA> 2015-06-01
3 2015-06-02 16:00:00 28.0       NA  <NA> 2015-06-02
4 2015-06-03 16:00:00 28.0       NA  <NA> 2015-06-03
5 2015-06-03 17:00:00 30.0       60  <NA> 2015-06-03
6 2015-06-05 07:00:00 23.0       NA  <NA> 2015-06-05

为什么aggregate() 改变了t 的结构？有人可以解释我缺少什么吗？

作为参考，有dput() 格式的样本数据集（有另一个变量Tim (difftime) 保存从一天开始的测量偏移量）：

> dput(t)
structure(list(Time = structure(list(sec = c(0, 0, 0, 0, 0, 0
), min = c(0L, 0L, 0L, 0L, 0L, 0L), hour = c(16L, 15L, 16L, 16L, 
17L, 7L), mday = c(18L, 1L, 2L, 3L, 3L, 5L), mon = c(4L, 5L, 
5L, 5L, 5L, 5L), year = c(115L, 115L, 115L, 115L, 115L, 115L), 
    wday = c(1L, 1L, 2L, 3L, 3L, 5L), yday = c(137L, 151L, 152L, 
    153L, 153L, 155L), isdst = c(1L, 1L, 1L, 1L, 1L, 1L), zone = c("CEST", 
    "CEST", "CEST", "CEST", "CEST", "CEST"), gmtoff = c(NA_integer_, 
    NA_integer_, NA_integer_, NA_integer_, NA_integer_, NA_integer_
    )), .Names = c("sec", "min", "hour", "mday", "mon", "year", 
"wday", "yday", "isdst", "zone", "gmtoff"), class = c("POSIXlt", 
"POSIXt")), Temp = c(26.5, 26.5, 28, 28, 30, 23), Humidity = c(NA, 
NA, NA, NA, 60, NA), Notes = c(NA_character_, NA_character_, 
NA_character_, NA_character_, NA_character_, NA_character_), 
    Day = structure(list(sec = c(0, 0, 0, 0, 0, 0), min = c(0L, 
    0L, 0L, 0L, 0L, 0L), hour = c(0L, 0L, 0L, 0L, 0L, 0L), mday = c(18L, 
    1L, 2L, 3L, 3L, 5L), mon = c(4L, 5L, 5L, 5L, 5L, 5L), year = c(115L, 
    115L, 115L, 115L, 115L, 115L), wday = c(1L, 1L, 2L, 3L, 3L, 
    5L), yday = c(137L, 151L, 152L, 153L, 153L, 155L), isdst = c(-1L, 
    -1L, -1L, -1L, -1L, -1L), zone = c("CEST", "CEST", "CEST", 
    "CEST", "CEST", "CEST"), gmtoff = c(NA_integer_, NA_integer_, 
    NA_integer_, NA_integer_, NA_integer_, NA_integer_)), .Names = c("sec", 
    "min", "hour", "mday", "mon", "year", "wday", "yday", "isdst", 
    "zone", "gmtoff"), class = c("POSIXlt", "POSIXt")), Tim = structure(c(16, 
    15, 16, 16, 17, 7), class = "difftime", units = "hours")), .Names = c("Time", 
"Temp", "Humidity", "Notes", "Day", "Tim"), row.names = c(NA, 
6L), class = "data.frame")

Answer 1

以下是使用dplyr、data.table 和split-lapply 组合的三种解决方案，无需任何封装：

在每种情况下，将Time 转换为POSIXct 和Day 转换为Date（我在此处调用了您提供的样本数据集并选择 27 作为截止值，因此我们有一天有两个匹配的行。）：

sample$Time <- as.POSIXct(sample$Time)
sample$Day  <- as.Date(sample$Day)

使用dplyr，动词不言自明，这就是为什么这是我最喜欢的解决方案：

require(dplyr)

result <- sample %>% 
          group_by(Day) %>% 
          summarise(greater27=max(Temp > 27))

result
# # A tibble: 5 x 2
#          Day greater27
#       <date>     <int>
# 1 2015-05-18         0
# 2 2015-06-01         0
# 3 2015-06-02         1
# 4 2015-06-03         1
# 5 2015-06-05         0

使用data.table，将第一个参数留空以选择所有行，在第二个参数中进行计算，并将by指定为命名参数：

require(data.table)

sample <- data.table(sample)
result <- sample[, .(greater27=max(Temp > 27)), by="Day"]

result
#           Day greater27
# 1: 2015-05-18         0
# 2: 2015-06-01         0
# 3: 2015-06-02         1
# 4: 2015-06-03         1
# 5: 2015-06-05         0

仅使用基础 R，split Day 的数据集，这将为您提供 data.frames 的列表，然后应用匿名函数，最后将 rbind 全部重新组合到一个 data.frame：

result <- do.call(rbind, 
                  lapply(split(sample, sample$Day),
                         function(x){
                           data.frame(
                             Day = x$Day[1],
                             greater27 = max(x$Temp > 27)
                           )
                         }
                    )
            )

result
#                   Day greater27
# 2015-05-18 2015-05-18         0
# 2015-06-01 2015-06-01         0
# 2015-06-02 2015-06-02         1
# 2015-06-03 2015-06-03         1
# 2015-06-05 2015-06-05         0