r - 如何返回相隔至少 10 天的每年的每月最小值

Question

我有一个每日时间序列，我想获得每年每个月的最小值，但我想确保结果至少相隔 10 天。更具体地说，让我们解释以下示例数据帧。

>Data
        Years   Months     Days Date        A   B
    1   2003    December    1   2003-12-01  10  10
    2   2003    December    2   2003-12-02  10  10
    3   2003    December    3   2003-12-03  10  10
    4   2003    December    4   2003-12-04  10  10
    5   2003    December    5   2003-12-05  10  10
    6   2003    December    6   2003-12-06  10  10
    7   2003    December    7   2003-12-07  10  10
    8   2003    December    8   2003-12-08  3   10
    9   2003    December    9   2003-12-09  10  10
    10  2003    December    10  2003-12-10  10  10
    11  2003    December    11  2003-12-11  10  10
    12  2003    December    12  2003-12-12  10  4
    13  2003    December    13  2003-12-13  10  10
    14  2003    December    14  2003-12-14  10  10
    15  2003    December    15  2003-12-15  10  10
    16  2003    December    16  2003-12-16  10  10
    17  2003    December    17  2003-12-17  10  10
    18  2003    December    18  2003-12-18  10  10
    19  2003    December    19  2003-12-19  10  10
    20  2003    December    20  2003-12-20  10  10
    21  2003    December    21  2003-12-21  10  10
    22  2003    December    22  2003-12-22  10  10
    23  2003    December    23  2003-12-23  10  10
    24  2003    December    24  2003-12-24  10  10
    25  2003    December    25  2003-12-25  10  10
    26  2003    December    26  2003-12-26  10  10
    27  2003    December    27  2003-12-27  10  10
    28  2003    December    28  2003-12-28  10  10
    29  2003    December    29  2003-12-29  10  10
    30  2003    December    30  2003-12-30  10  10
    31  2003    December    31  2003-12-31  10  10
    32  2004    January     1   2004-01-01  10  10
    33  2004    January     2   2004-01-02  10  10
    34  2004    January     3   2004-01-03  10  10
    35  2004    January     4   2004-01-04  10  10
    36  2004    January     5   2004-01-05  10  10
    37  2004    January     6   2004-01-06  10  10
    38  2004    January     7   2004-01-07  10  10
    39  2004    January     8   2004-01-08  10  10
    40  2004    January     9   2004-01-09  10  10
    41  2004    January     10  2004-01-10  10  10
    42  2004    January     11  2004-01-11  10  10
    43  2004    January     12  2004-01-12  10  10
    44  2004    January     13  2004-01-13  10  10
    45  2004    January     14  2004-01-14  10  10
    46  2004    January     15  2004-01-15  10  10
    47  2004    January     16  2004-01-16  10  10
    48  2004    January     17  2004-01-17  10  10
    49  2004    January     18  2004-01-18  10  10
    50  2004    January     19  2004-01-19  10  10
    51  2004    January     20  2004-01-20  10  10
    52  2004    January     21  2004-01-21  10  10
    53  2004    January     22  2004-01-22  10  10
    54  2004    January     23  2004-01-23  10  10
    55  2004    January     24  2004-01-24  10  10
    56  2004    January     25  2004-01-25  5   4
    57  2004    January     26  2004-01-26  10  10
    58  2004    January     27  2004-01-27  10  10
    59  2004    January     28  2004-01-28  10  10
    60  2004    January     29  2004-01-29  10  10
    61  2004    January     30  2004-01-30  10  10
    62  2004    January     31  2004-01-31  10  10
    63  2004    February    1   2004-02-01  10  10
    64  2004    February    2   2004-02-02  5   4
    65  2004    February    3   2004-02-03  10  10
    66  2004    February    4   2004-02-04  10  10
    67  2004    February    5   2004-02-05  10  10
    68  2004    February    6   2004-02-06  10  10
    69  2004    February    7   2004-02-07  10  10
    70  2004    February    8   2004-02-08  10  10
    71  2004    February    9   2004-02-09  7   6
    72  2004    February    10  2004-02-10  10  10
    73  2004    February    11  2004-02-11  10  10
    74  2004    February    12  2004-02-12  10  10
    75  2004    February    13  2004-02-13  10  10
    76  2004    February    14  2004-02-14  10  10
    77  2004    February    15  2004-02-15  10  10
    78  2004    February    16  2004-02-16  10  10
    79  2004    February    17  2004-02-17  10  10
    80  2004    February    18  2004-02-18  10  10
    81  2004    February    19  2004-02-19  10  10
    82  2004    February    20  2004-02-20  10  10
    83  2004    February    21  2004-02-21  10  10
    84  2004    February    22  2004-02-22  10  10
    85  2004    February    23  2004-02-23  10  10
    86  2004    February    24  2004-02-24  10  10
    87  2004    February    25  2004-02-25  10  10
    88  2004    February    26  2004-02-26  10  10
    89  2004    February    27  2004-02-27  10  10
    90  2004    February    28  2004-02-28  10  10
    91  2004    February    29  2004-02-29  10  10

我想做的几乎是aggregate()所做的

min <- aggregate(Data[5:6], by= list(Data$Months, Data$Years), FUN = min)


Group.1     Group.2 A   B
December    2003    3   4
January     2004    5   4
February    2004    5   4

但是，要在 2 月获得与前几个月的最小值相距至少 10 天的每个 A 和 B 的最小值。

所以我想得到:

Group.1     Group.2 A   B
December    2003    3   4
January     2004    5   4
February    2004    7   6

有任何想法吗？

Answer 1

好吧，如果您有兴趣，我有一个凌乱的解决方案:)

首先，让我们确保月份正确排序并为月份/年份组合创建一个因子

data$Months<-factor(data$Months, levels=month.name)
data$MY<-interaction(data$Months, data$Years, drop=T)

现在我将定义一些辅助函数

getpaddoff<-function(n) {
    function(x) {
        a<-which.min(x)+n-length(x); 
        ifelse(a>0,a,0) 
    }
}
rollright<-function(x, add=0) {
  n<-names(x)
  x<-head(c(add,x), -1)
  names(x)<-n;
  x
}

getpadoff函数将返回下个月有多少所需的非重叠天数。和 rollright将允许我将返回从一个月转移到下一个月。 getpadoff正如所写的那样，它要求每个月的每一天在数据中都有一个条目。

好的，现在我们开始将这些应用于数据。我们得到一个函数来确保 10 天的间隔。然后我们根据月/年拆分数据。然后我们计算我们必须从每个月移除的天数，因为最小值太接近上个月的月底了。

paddoff <- getpaddoff(10)
ss <- split(data[c("A","B")], data$MY)
offsets <- rollright(lapply(ss, function(x) sapply(x, padoff)), 
    add=list(c(A=0, B=0)))

一旦我们有了这些值，我们就可以找到每个月的非重叠最小值。

rr<-Map(function(d,off) {
    d<-as.matrix(d)
    stopifnot(ncol(d)==length(off))
    for(i in seq_along(off)) {
         if(off[i]>0)
             d[1:off[i],i]<-Inf
    }
    apply(d,2,min)
}, ss, offsets)
do.call(rbind,rr)

这是结果

              A B
December.2003 3 4
January.2004  5 4
February.2004 7 6

我不确定您究竟需要如何格式化结果，但这至少可以提取您想要的值。