在矩阵上进行最小/最大聚合后获取位置索引

Question

Sequel 我的第一个问题 ()。现在我想计算每个 12 小时间隔的时间序列的最大值（和最小值）。

我得到了每小时的数据测量值 (data_measure)。现在改成了半天的时间序列

t_measure <- ts(data = data_measure, frequency = 12)

然后我使用了来自{stats}的聚合函数

data_measure_daily_max <- aggregate(t_measure, 1, max)

data_measure <- structure(c(8.29, 7.96, 8.14, 7.27, 7.37, 7.3, 7.23, 7.53, 
7.98, 10.2, 12.39, 14.34, 14.87, 14.39, 12.54, 11.84, 10.3, 10.62, 
10.65, 10.56, 10.43, 10.35, 9.85, 9.12, 8.95, 8.82, 8.92, 9.33, 
9.44, 9.3, 9.15, 9.37, 9.54, 10.24, 12.13, 12.43, 12.65, 13, 
13.18, 13.58, 13.64, 13.75, 13.85, 13.94, 13.79, 13.84, 13.94, 
14.26, 24.93, 24.64, 23.67, 21.46, 21.33, 20.83, 21.12, 21.1, 
23.75, 25.39, 30.72, 30.71, 30.81, 30.92, 32.61, 32.37, 32.49, 
30.68, 30.23, 30.45, 28.1, 26.9, 25.09, 25.07, 24.59, 24.22, 
23.05, 22.21, 22.07, 21.6, 21.24, 21.22, 21.85, 24.87, 28.85, 
29.42, 30.82, 30.97, 31.32, 30.81, 30.83, 29.9, 30.01, 30.31, 
30, 27.91, 25.78, 25.88, 8.78, 8.47, 8.49, 7.65, 8.63, 9.02, 
9.02, 8.11, 7.63, 9.19, 11.25, 12.24, 13.62, 12.09, 10.6, 11.1, 
10.16, 10.44, 9.58, 10.04, 10.01, 10.23, 9.51, 9.2, 9.34, 9.6, 
9.4, 9.45, 9.36, 9.26, 9.3, 9.46, 9.58, 9.89, 10.6, 11.04, 12.1, 
12.61, 13.12, 13.47, 13.55, 13.51, 13.63, 13.84, 13.93, 14.17, 
13.97, 13.86), .Dim = c(48L, 3L), .Dimnames = list(NULL, c("station1", 
"station2", "station3")))

所以实际上我需要一个 index/vector 来告诉我这些时间间隔的最大值和最小值在哪里，所以稍后我可以准确地提取这些时间间隔以供其他数据集进行比较。

我的第一次试用：

max_index <- which(aggregate(t_measure, 1, max))   # argument to 'which' is not logical

Answer 1

将 which.max 和 which.min 与 aggregate

结合使用

a1 <- aggregate(t_measure, 1, which.min)
a2 <- aggregate(t_measure, 1, which.max)

a1
#Time Series:
#Start = 1 
#End = 4 
#Frequency = 1 
#  station1 station2 station3
#1        7        6        9
#2       12       12       12
#3        2        8        6
#4        1       11        1

a2
#Time Series:
#Start = 1 
#End = 4 
#Frequency = 1 
#  station1 station2 station3
#1       12       11       12
#2        1        3        1
#3       12       12       12
#4       12        3       10

如果您想要参考原始 data_measure 数据帧的最小索引，我们可以做到

vals <- nrow(t_measure)/12 
index_min <- a1 + (12 * (seq_len(vals) - 1))
index_min

#Time Series:
#Start = 1 
#End = 4 
#Frequency = 1 
#  station1 station2 station3
#1        7        6        9
#2       24       24       24
#3       26       32       30
#4       37       47       37

这可以理解为 station1 在第一个 12 小时间隔中最大值出现在 data_measure 的第 7 行中，对于接下来的 12 小时间隔，它出现在第 24 行中，其他相同站。

在矩阵上进行最小/最大聚合后获取位置索引

Get position indices after min / max aggregation on matrix

aggregate

r

max

min