我可以使用数据表连接或聚合函数来聚合开启日期和关闭日期之间的时间序列数据吗？

Question

我想有效地总结收集离散样本期间的连续气象数据。

我目前使用一个耗时的循环来执行此操作，但我想存在更好的解决方案。我是 data.table 语法的新手，但似乎应该有一个加入的解决方案。

continuous <- data.frame(Time = seq(as.POSIXct("2019-01-01 0:00:00"),
                                    as.POSIXct("2019-01-01 9:00:00"),"hour"),
                         CO2  = sample(400:450,10), 
                         Temp = sample(10:30,10))

> continuous
                  Time CO2 Temp
1  2019-01-01 00:00:00 430   11
2  2019-01-01 01:00:00 412   26
3  2019-01-01 02:00:00 427   17
4  2019-01-01 03:00:00 435   29
5  2019-01-01 04:00:00 447   23
6  2019-01-01 05:00:00 417   19
7  2019-01-01 06:00:00 408   12
8  2019-01-01 07:00:00 449   28
9  2019-01-01 08:00:00 445   20
10 2019-01-01 09:00:00 420   27

discrete <- data.frame(on = c(as.POSIXct("2019-01-01 0:00:00"),
                              as.POSIXct("2019-01-01 3:00:00")),
                      off = c(as.POSIXct("2019-01-01 3:00:00"),
                              as.POSIXct("2019-01-01 8:00:00")))

> discrete
                   on                 off
1 2019-01-01 00:00:00 2019-01-01 03:00:00
2 2019-01-01 03:00:00 2019-01-01 08:00:00

discrete[, c("CO2.mean","Temp.mean")] <- 
    lapply(seq(length(c("CO2","Temp"))), function(k) 
        unlist(lapply(seq(length(discrete[, 1])), function(i) 
            mean(continuous[
                which.closest(continuous$Time,discrete$on[i]):
                which.closest(continuous$Time, discrete$off[i]), 
                    c("CO2","Temp")[k]]))))

> discrete
                   on                 off CO2.mean Temp.mean
1 2019-01-01 00:00:00 2019-01-01 03:00:00    426.0  20.75000
2 2019-01-01 03:00:00 2019-01-01 08:00:00    433.5  21.83333

这是可行的，但是当将数十个连续变量聚合成数百个采样周期时，需要很长时间才能运行。感谢您的帮助！

Answer 1

一个选项是data.table中的非等连接

library(data.table)
setDT(continuous)[discrete, .(CO2mean = mean(CO2), 
    Tempmean = mean(Temp)),on = .(Time >= on, Time <= off), by = .EACHI]

---

或使用滚动连接

setDT(continuous)[discrete, .(CO2mean = mean(CO2), 
   Tempmean = mean(Temp)),on = .(Time = on, Time = off),
        by = .EACHI, roll = 'nearest']