Question

我有一个数据集，在其他列中有date, sequence and low列，请参见下面的df。 1-to-9列中的序列sequence被视为一个块或一个完整周期数据集具有几个这样的完整块/循环和部分完成的{/ {1}}

这就是我要解决的问题：

删除部分完成的周期，然后对整个周期进行分组（请参见eg: 1-to-4）
对于每个块/周期（即从1到9的序列），我想找到低点以及低点发生的日期。

如果存在两个具有相同值但在不同日期的低点，则它应该只输出最新日期（请参见输出中的第三个方框）

df1

按周期/块分组的数据

library(lubridate)
library(tidyverse)
### Sample data
df <- data.frame(stringsAsFactors=FALSE,
date = c("1/01/2019", "2/01/2019", "3/01/2019", "4/01/2019",
"5/01/2019", "6/01/2019", "7/01/2019", "8/01/2019",
"9/01/2019", "10/01/2019", "11/01/2019", "12/01/2019", "13/01/2019",
"14/01/2019", "15/01/2019", "16/01/2019", "17/01/2019", "18/01/2019",
"19/01/2019", "20/01/2019", "21/01/2019", "22/01/2019",
"23/01/2019", "24/01/2019", "25/01/2019", "26/01/2019", "27/01/2019",
"28/01/2019", "29/01/2019", "30/01/2019", "31/01/2019",
"1/02/2019", "2/02/2019", "3/02/2019", "4/02/2019"),
sequence = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 1, 2, 3, 4, 5, 6, 7, 8,
9, 1, 2, 3, 4, 1, 2, 3, 4, 5, 6, 7, 8, 9),
low = c(96, 81, 43, 18, 43, 65, 48, 90, 69, 50, 41, 73, 1, 1, 7, 49,
16, 79, 2, 74, 8, 88, 56, 57, 66, 29, 79, 51, 52, 47, 42, 9,
41, 9, 50)) %>% mutate(date = dmy(date))

我要的最终输出

df1 <- data.frame(stringsAsFactors=FALSE,
    date = c("1/01/2019", "2/01/2019", "3/01/2019", "4/01/2019",
             "5/01/2019", "6/01/2019", "7/01/2019", "8/01/2019",
             "9/01/2019", "14/01/2019", "15/01/2019", "16/01/2019", "17/01/2019",
             "18/01/2019", "19/01/2019", "20/01/2019", "21/01/2019", "22/01/2019",
             "27/01/2019", "28/01/2019", "29/01/2019", "30/01/2019",
             "31/01/2019", "1/02/2019", "2/02/2019", "3/02/2019", "4/02/2019"),
sequence = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3,
             4, 5, 6, 7, 8, 9),
     low = c(96, 81, 43, 18, 43, 65, 48, 90, 69, 1, 7, 49, 16, 79, 2, 74,
             8, 88, 79, 51, 52, 47, 42, 9, 41, 9, 50),
   group = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3,
             3, 3, 3, 3, 3, 3)) %>% mutate(date = dmy(date))

有什么想法吗？
附言我在格式化这个问题时遇到了一些问题，因此很不统一。

Answer 1

我们通过取序列为1的累积总和来创建分组变量，然后filter仅包含9个元素的组，而slice在{{1之后， }}按arrange结尾的顺序排列“日期”，以照顾到“最低”价值存在联系的情况

desc

或df %>% group_by(group = cumsum(sequence == 1)) %>% filter(n() == 9) %>% select(date, low) %>% arrange(desc(date)) %>% slice(which.min(low)) %>% ungroup %>% select(-group) # A tibble: 3 x 2 # date low # <date> <dbl> #1 2019-01-04 18 #2 2019-01-14 1 #3 2019-02-03 9的类似选项

data.table

Answer 2

另一种dplyr可能是：

df %>%
 group_by(group = cumsum(sequence == 1), rleid = with(rle(group), rep(seq_along(lengths), lengths))) %>%
 filter(all(c(1:9) %in% sequence)) %>%
 slice(which.min(rank(low, ties.method = "last"))) %>%
 ungroup() %>%
 select(-group, -rleid)

  date       sequence   low
  <date>        <dbl> <dbl>
1 2019-01-04        4    18
2 2019-01-14        1     1
3 2019-02-03        8     9

在这里，首先，创建“ sequence” == 1的累积和以及基于该累积和的rleid()类变量，然后按两者进行分组。其次，它消除了序列不包含所有九个值的情况。最后，在关系返回最后一个最小值的情况下，它返回每组的最小值（您可以通过参数ties.method对其进行修改。）

Answer 3

在基数R中也是可能的。不过，可能有点 map sy。

w <- which(df$sequence == 1)
w <- w[sapply(w, function(x) df$sequence[x + 8] == 9 & sum(df$sequence[x:(x + 8)]) == 45)]
do.call(rbind, Map(function(x) x[which.min(x$low), ], 
                   Map(function(s) df[s, ], Map(seq, w, l=9))))
#          date sequence low
# 4  2019-01-04        4  18
# 14 2019-01-14        1   1
# 32 2019-02-01        6   9

诀窍是找到完成的序列并将它们分组在列表中，然后rbind which.min每组。如果实际上没有错误的序列，则应该考虑sum(.) == 45检查。

数据

df <- structure(list(date = structure(c(17897, 17898, 17899, 17900, 
17901, 17902, 17903, 17904, 17905, 17906, 17907, 17908, 17909, 
17910, 17911, 17912, 17913, 17914, 17915, 17916, 17917, 17918, 
17919, 17920, 17921, 17922, 17923, 17924, 17925, 17926, 17927, 
17928, 17929, 17930, 17931), class = "Date"), sequence = c(1, 
2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 1, 2, 3, 4, 5, 6, 7, 8, 9, 
1, 2, 3, 4, 1, 2, 3, 4, 5, 6, 7, 8, 9), low = c(96, 81, 43, 18, 
43, 65, 48, 90, 69, 50, 41, 73, 1, 1, 7, 49, 16, 79, 2, 74, 8, 
88, 56, 57, 66, 29, 79, 51, 52, 47, 42, 9, 41, 9, 50)), row.names = c(NA, 
-35L), class = "data.frame")

按顺序分组，然后在列中找到最小值

按周期/块分组的数据

3 个答案: