不理解R中的滞后和窗口

Question

我有点搞不懂滞后和窗户是怎么工作的。从概念上看，它似乎很简单:延迟基本上是通过“行”数将数据前后移动，而窗口就像一个“select”操作，根据开始和结束时间来指示要使用的数据(据我所理解)。

以下是“用R进行时间序列分析”中的一些代码

library(TSstudio)
data(USVSales)
lags  <- function(ts.obj, l) {
ts_merged <- NULL
## Creating n lags
for (i in 1:l) {
    ## So, for each lag in 1:l, this adds the ts.obj, lagged by i
    ## to ts_merged, so by the end of this loop we have
    ## ts(lag=-1), ts(lag=-2) ..., ts(lag=-l)
    ts_merged <- ts.union(ts_merged, stats::lag(ts.obj,k=-i))
}
## Merge with the original series (adds the original)
ts_merged <- ts.union(ts.obj,ts_merged)
## set the column names (but how do we know that the first
## 'column' is the original?)
colnames(ts_merged <- c("y",paste0("y_",1:i)))
## Removing NAs as results of lag creation
ts_merged <- window(ts_merged,
                    start=start(ts.obj)+l, 
                    end=end(ts.obj))
return(ts_merged)
}

我制作了两个不同版本的延迟函数，一个基于图书代码(其结果将命名为' book ')另一个相同，但调用window()时注释掉(输出名为'windowless')

书中的例子继续下去..。

> book <- book_lags(USVSales,l=3)
> head(book)
          y    y_1    y_2    y_3
[1,] 1304.6 1483.8 1148.7 1077.5
[2,] 1373.0 1304.6 1483.8 1148.7
[3,] 1183.8 1373.0 1304.6 1483.8
[4,] 1164.4 1183.8 1373.0 1304.6
[5,] 1207.2 1164.4 1183.8 1373.0
[6,] 1029.6 1207.2 1164.4 1183.8

我的代码生成以下内容

> windowless <- windowless_lags(USVSales,l=3)
> head(windowless)
      y    y_1    y_2    y_3
[1,]  885.2     NA     NA     NA
[2,]  994.7  885.2     NA     NA
[3,] 1243.6  994.7  885.2     NA
[4,] 1191.2 1243.6  994.7  885.2
[5,] 1203.2 1191.2 1243.6  994.7
[6,] 1254.7 1203.2 1191.2 1243.6

如果然后将对窗口的调用应用于“无窗口”输出，则得到与书中相同的输出。

head(window(windowless,start=start(windowless)+3,
 end=end(windowless)))
      y    y_1    y_2    y_3
[1,] 1304.6 1483.8 1148.7 1077.5
[2,] 1373.0 1304.6 1483.8 1148.7
[3,] 1183.8 1373.0 1304.6 1483.8
[4,] 1164.4 1183.8 1373.0 1304.6
[5,] 1207.2 1164.4 1183.8 1373.0
[6,] 1029.6 1207.2 1164.4 1183.8

我不明白为什么第一排“(头(没有窗户)，.”不等于“无窗口”的第四行。1304.6是从哪里来的？

我遗漏了什么？

Answer 1

神秘的是，window()中的‘`start’关键字可以给出一个数字，但是(在本例中)它被解释为，是年增量和月增量。

> foo <- ts(data=1:60,frequency=12)
> ts_info(foo)
The foo series is a ts object with 1 variable and 60  observations
Frequency: 12 
Start time: 1 1 
End time: 5 12 

> bar <- window(foo,start=start(foo)+3,end=end(foo))
> ts_info(bar)

The bar series is a ts object with 1 variable and 21 observations

Frequency: 12 
Start time: 4 4 
End time: 5 12 

“1304.6”来自于USVSales ts对象第三年的第三个月

> ts_info(USVSales)
 The USVSales series is a ts object with 1 variable and 528 observations
 Frequency: 12 
 Start time: 1976 1 
 End time: 2019 12
 
> baz <- window(USVSales,start=start(USVSales)+3,end=start(USVSales)+6)
> ts_info(baz)

 The baz series is a ts object with 1 variable and 40 observations
 Frequency: 12 
 Start time: 1979 4 
 End time: 1982 7 
> 

head(baz,1)
[1] 1304.6