我有 5 个地点 15 天的时间序列数据(以日格式)存储为matrix
。数据的结构是
meter_daywise<-structure(c(24.4745528484842, 21.5936510486629, 58.9120896540103,
49.4188338105575, 568.791971631185, 27.1682608244523, 23.3482757939878,
74.710966227615, 82.6947717673258, 704.212340152625, 23.7581651139442,
21.154634543401, 64.9680107059625, 420.903181621575, 672.629513512841,
128.22871420984, 601.521395359887, 74.6606087800009, 335.87599588534,
576.451039365565, 641.329910104503, 1010.78497435794, 72.6159099850862,
225.153924410613, 582.652388366075, 529.082673064516, 1151.87208010484,
76.9939865858514, 198.567927906582, 641.511944831027, 280.685806121688,
998.647413766557, 73.2033388656998, 337.966543898629, 847.24874747014,
76.7357959402453, 1065.75153722813, 220.286408574643, 301.120955096701,
552.703945876515, 206.496034127105, 1053.49582469841, 206.187963352323,
219.791668265415, 655.496754449233, 172.87981151456, 1018.01514547636,
544.551001017031, 227.116788647859, 656.566145328213, 373.484460701849,
1503.65562864399, 117.732932835236, 251.383369528816, 802.871808716031,
150.471195301885, 1414.88799728991, 14.6490905509617, 203.429955747521,
622.731792495107, 548.093577186778, 1076.5618643676, 15.5135269483705,
256.581499048612, 644.572474965446, 63.2304035656636, 1538.07906461011,
15.0980567507389, 261.513768642083, 622.17970609429, 210.786387991582,
996.998005580537, 15.8138368515615, 157.390773346978, 573.477606081416
), .Dim = c(5L, 15L), .Dimnames = list(c("apFac_401", "apFac_403",
"apFac_501", "apFac_503", "apFac_601"), c("D1", "D2", "D3", "D4",
"D5", "D6", "D7", "D8", "D9", "D10", "D11", "D12", "D13", "D14",
"D15")))
早些时候,我使用以下方法计算不同系列之间的相关性
library(corrplot)# for plotting correlation matrix
corrplot(cor(t(meter_daywise)),method = "number",type="lower")# have taken transpose of above structure
So, with this I am getting a nice correlation matrix showing correlation between different series.
但是,在观察相关值时,我发现有些问题,在搜索时我发现了这个link https://stats.stackexchange.com/questions/29096/correlation-between-two-time-series,其中提到我们需要计算互相关。因此,现在我需要像上面那样计算互相关矩阵。因此,我发现了一些函数,例如
1. ccf() #in base packages
2. diss(meter_daywise,METHOD = "CORT",deltamethod = "DTW")#in TSclust package
我在使用上述功能时遇到两个问题:
-
ccf
不将完整矩阵作为输入
-
diss()
接受输入矩阵并生成一些矩阵,但是在观察这些值时,我发现它不是互相关矩阵,因为这些值不在之间-1
and 1
.
那么问题是我们如何计算R中不同时间序列值的互相关矩阵?