堆栈内存溢出
  简体   繁体   English

计算 dplyr 中两组不同列的 group_by() 中的累积和

[英]Calculate cumulative sum in a group_by() on two different sets of columns in dplyr

 原文  2020-06-12 13:18:28  8   5    r/ dplyr/ tidyverse

问题描述

My initial dataframe looks like:我最初的 dataframe 看起来像：

library(tidyverse)

df_input <- data.frame(
            cohort = c("2019-03-01", "2019-03-01", "2019-03-01", "2019-03-01",
                       "2019-03-01", "2019-03-01", "2019-03-01", "2019-03-01",
                       "2019-03-01", "2019-04-01", "2019-04-01", "2019-04-01",
                       "2019-04-01", "2019-04-01", "2019-04-01", "2019-04-01"),
            months = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 6, 7),
               CLV = c(59.9, 61.1, 62.06, 62.58, 62.83, NA, NA, NA, NA, 22.2, 38.24,
                       46.08, 56.28, NA, NA, NA),
           CLV_for = c(1.66, 1.42, 1.42, 1.42, 1.18, 1.18, 1.18, 1.18, 0.95, 35.75,
                       26.1, 16.09, 10.37, 7.15, 6.08, 5.01)
      )

       cohort months   CLV CLV_for
1  2019-03-01      1 59.90    1.66
2  2019-03-01      2 61.10    1.42
3  2019-03-01      3 62.06    1.42
4  2019-03-01      4 62.58    1.42
5  2019-03-01      5 62.83    1.18
6  2019-03-01      6    NA    1.18
7  2019-03-01      7    NA    1.18
8  2019-03-01      8    NA    1.18
9  2019-03-01      9    NA    0.95
10 2019-04-01      1 22.20   35.75
11 2019-04-01      2 38.24   26.10
12 2019-04-01      3 46.08   16.09
13 2019-04-01      4 56.28   10.37
14 2019-04-01      5    NA    7.15
15 2019-04-01      6    NA    6.08
16 2019-04-01      7    NA    5.01

I want to perform a cumulative sum (using cumsum() in dplyr ) starting from the last non-NA value in each group (aka cohort ) in column CLV and continuing for the remaining correspondent values in the column CLV_for .我想从CLV列中每个组（又名cohort ）中的最后一个非 NA 值开始执行累积和（在dplyr中使用cumsum() ，然后继续执行CLV_for列中剩余的对应值。

In order to better exaplain the calculation, I thought of splitting it in 2 different steps.为了更好地解释计算，我想把它分成两个不同的步骤。

1) Starting from the last non-NA value in CLV column for cohort 2019-03-01 , cumsum() the corresponding values in column CLV_for . 1) 从队列2019-03-01的 CLV 列中的最后一个非 NA 值开始， cumsum()列CLV_for中的相应值。 Same for the cohort 2019-04-01 .队列2019-04-01 。

 df_inter <- data.frame(
  cohort = c("2019-03-01", "2019-03-01", "2019-03-01", "2019-03-01",
             "2019-03-01", "2019-03-01", "2019-03-01", "2019-03-01",
             "2019-03-01", "2019-04-01", "2019-04-01", "2019-04-01",
             "2019-04-01", "2019-04-01", "2019-04-01", "2019-04-01"),
  months = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 6, 7),
  CLV = c(59.9, 61.1, 62.06, 62.58, 62.83, NA, NA, NA, NA, 22.2, 38.24,
          46.08, 56.28, NA, NA, NA),
  cum_CLV_for = c(NA, NA, NA, NA, NA, 64.01, 65.19, 66.37, 67.32, NA,
                  NA, NA, NA, 63.43, 69.51, 74.51)
)

       cohort months   CLV cum_CLV_for
1  2019-03-01      1 59.90          NA
2  2019-03-01      2 61.10          NA
3  2019-03-01      3 62.06          NA
4  2019-03-01      4 62.58          NA
5  2019-03-01      5 62.83          NA
6  2019-03-01      6    NA       64.01 (<- 62.83 + 1.18)
7  2019-03-01      7    NA       65.19 (<- 64.01 + 1.18)
8  2019-03-01      8    NA       66.37 (<- 65.19 + 1.18)
9  2019-03-01      9    NA       67.32 (<- 66.37 + 0.95)
10 2019-04-01      1 22.20          NA
11 2019-04-01      2 38.24          NA
12 2019-04-01      3 46.08          NA
13 2019-04-01      4 56.28          NA
14 2019-04-01      5    NA       63.43 (<- 56.28 + 7.15)
15 2019-04-01      6    NA       69.51 (<- 63.43 + 6.08)
16 2019-04-01      7    NA       74.51 (<- 69.51 + 5.01)

2) The second step is to clean out the two columns merging them into one. 2）第二步是清理将它们合并为一的两列。

df_final <- data.frame(
                                      sub_date = c("2019-03-01", "2019-03-01", "2019-03-01",
                                                   "2019-03-01", "2019-03-01",
                                                   "2019-03-01", "2019-03-01",
                                                   "2019-03-01", "2019-03-01",
                                                   "2019-04-01", "2019-04-01",
                                                   "2019-04-01", "2019-04-01",
                                                   "2019-04-01", "2019-04-01",
                                                   "2019-04-01"),
                      months_after_acquisition = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 6, 7),
                                       cum_CLV = c(59.9, 61.1, 62.06, 62.58, 62.83, 64.01, 65.19,
                                                   66.37, 67.32, 22.2, 38.24,
                                                   46.08, 56.28, 63.43, 69.51,
                                                   74.51)
                   )

     sub_date months_after_acquisition cum_CLV
1  2019-03-01                        1   59.90
2  2019-03-01                        2   61.10
3  2019-03-01                        3   62.06
4  2019-03-01                        4   62.58
5  2019-03-01                        5   62.83
6  2019-03-01                        6   64.01
7  2019-03-01                        7   65.19
8  2019-03-01                        8   66.37
9  2019-03-01                        9   67.32
10 2019-04-01                        1   22.20
11 2019-04-01                        2   38.24
12 2019-04-01                        3   46.08
13 2019-04-01                        4   56.28
14 2019-04-01                        5   63.43
15 2019-04-01                        6   69.51
16 2019-04-01                        7   74.51

Thanks for your help!谢谢你的帮助！

5 个解决方案

解决方案1
 2  2020-06-12 13:29:41

By taking either CLV or the vertically filled value of CLV combined with cumsum , we get what you want:通过将CLV或CLV的垂直填充值与cumsum相结合，我们得到你想要的：

df_input %>% 
  group_by(cohort) %>% 
  arrange(months, .by_group = T) %>% 
  mutate(cum_CLV = CLV) %>% 
  fill(cum_CLV) %>% 
  mutate(cum_CLV = cum_CLV + cumsum(CLV_for*is.na(CLV)))


#  cohort     months   CLV CLV_for cum_CLV
#    <fct>       <dbl> <dbl>   <dbl>   <dbl>
#  1 2019-03-01      1  59.9    1.66    59.9
#  2 2019-03-01      2  61.1    1.42    61.1
#  3 2019-03-01      3  62.1    1.42    62.1
#  4 2019-03-01      4  62.6    1.42    62.6
#  5 2019-03-01      5  62.8    1.18    62.8
#  6 2019-03-01      6  NA      1.18    64.0
#  7 2019-03-01      7  NA      1.18    65.2
#  8 2019-03-01      8  NA      1.18    66.4
#  9 2019-03-01      9  NA      0.95    67.3
# 10 2019-04-01      1  22.2   35.8     22.2
# 11 2019-04-01      2  38.2   26.1     38.2
# 12 2019-04-01      3  46.1   16.1     46.1
# 13 2019-04-01      4  56.3   10.4     56.3
# 14 2019-04-01      5  NA      7.15    63.4
# 15 2019-04-01      6  NA      6.08    69.5
# 16 2019-04-01      7  NA      5.01    74.5

解决方案2
 0  2020-06-12 13:28:49

Here's an approach with case_when :这是case_when的一种方法：

library(dplyr)
df_input %>% 
  group_by(cohort) %>%
  mutate(CumCLV = cumsum(case_when(is.na(CLV) ~ CLV_for,
                            TRUE ~ 0)),
         CLV = case_when(is.na(CLV) ~ CumCLV + max(CLV, na.rm = TRUE), 
                         TRUE ~ CLV)) %>%
  dplyr::select(-CLV_for, -CumCLV)

# A tibble: 16 x 3
# Groups:   cohort [2]
   cohort     months   CLV
   <fct>       <dbl> <dbl>
 1 2019-03-01      1  59.9
 2 2019-03-01      2  61.1
 3 2019-03-01      3  62.1
 4 2019-03-01      4  62.6
 5 2019-03-01      5  62.8
 6 2019-03-01      6  64.0
 7 2019-03-01      7  65.2
 8 2019-03-01      8  66.4
 9 2019-03-01      9  67.3
10 2019-04-01      1  22.2
11 2019-04-01      2  38.2
12 2019-04-01      3  46.1
13 2019-04-01      4  56.3
14 2019-04-01      5  63.4
15 2019-04-01      6  69.5
16 2019-04-01      7  74.5

解决方案3
 0  2020-06-12 13:45:09

a data.table approach for completeness sake为了完整起见，采用 data.table 方法

setDT(df_input)
df_input[, max := max(CLV, na.rm = TRUE), by = cohort]
df_input[ is.na(CLV), CLV := max + cumsum(CLV_for), by = cohort ][, c("max", "CLV_for") := NULL][]

#        cohort months   CLV
# 1: 2019-03-01      1 59.90
# 2: 2019-03-01      2 61.10
# 3: 2019-03-01      3 62.06
# 4: 2019-03-01      4 62.58
# 5: 2019-03-01      5 62.83
# 6: 2019-03-01      6 64.01
# 7: 2019-03-01      7 65.19
# 8: 2019-03-01      8 66.37
# 9: 2019-03-01      9 67.32
# 10: 2019-04-01      1 22.20
# 11: 2019-04-01      2 38.24
# 12: 2019-04-01      3 46.08
# 13: 2019-04-01      4 56.28
# 14: 2019-04-01      5 63.43
# 15: 2019-04-01      6 69.51
# 16: 2019-04-01      7 74.52

解决方案4
 0  2020-06-12 13:55:18

Yet another dplyr possibility could be:另一个dplyr可能性可能是：

df_input %>%
 group_by(cohort) %>%
 transmute(months,
           CLV = if_else(is.na(CLV), 
                         last(na.omit(CLV)) + cumsum(CLV_for * is.na(CLV)),
                         CLV))

   cohort     months   CLV
   <fct>       <dbl> <dbl>
 1 2019-03-01      1  59.9
 2 2019-03-01      2  61.1
 3 2019-03-01      3  62.1
 4 2019-03-01      4  62.6
 5 2019-03-01      5  62.8
 6 2019-03-01      6  64.0
 7 2019-03-01      7  65.2
 8 2019-03-01      8  66.4
 9 2019-03-01      9  67.3
10 2019-04-01      1  22.2
11 2019-04-01      2  38.2
12 2019-04-01      3  46.1
13 2019-04-01      4  56.3
14 2019-04-01      5  63.4
15 2019-04-01      6  69.5
16 2019-04-01      7  74.5

解决方案5
 0  2020-06-12 14:25:15

Using purrr::accumulate2() :使用purrr::accumulate2() ：

library(purrr)
library(dplyr)

df_input %>%
  group_by(cohort) %>%
  mutate(CLV = flatten_dbl(accumulate2(CLV, CLV_for[-1], .f = ~ if(!is.na(..2)) ..2 else ..1 + ..3))) %>%
  select(-CLV_for)

# A tibble: 16 x 3
# Groups:   cohort [2]
   cohort     months   CLV
   <chr>       <dbl> <dbl>
 1 2019-03-01      1  59.9
 2 2019-03-01      2  61.1
 3 2019-03-01      3  62.1
 4 2019-03-01      4  62.6
 5 2019-03-01      5  62.8
 6 2019-03-01      6  64.0
 7 2019-03-01      7  65.2
 8 2019-03-01      8  66.4
 9 2019-03-01      9  67.3
10 2019-04-01      1  22.2
11 2019-04-01      2  38.2
12 2019-04-01      3  46.1
13 2019-04-01      4  56.3
14 2019-04-01      5  63.4
15 2019-04-01      6  69.5
16 2019-04-01      7  74.5

暂无
暂无

声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

相关问题 R dplyr group_by 两个不同的列并总结 - R dplyr group_by two different columns and summarise dplyr group_by遍历不同的列 - dplyr group_by loop through different columns dplyr group_by两列的元素 - dplyr group_by over elements of two columns 在 dplyr 中，group_by() + summarise(sum) 不起作用 - In dplyr group_by() + summarise(sum)is not working 使用 dplyr group_by 计算总和的问题 - Problem with calculation of sum with usage of dplyr group_by 使用 dplyr 计算时间戳的累积总和 - Calculate cumulative sum over time stamp with dplyr dplyr使用group_by和rowwise do对累积集计数进行分组 - dplyr grouped cumulative set counting using group_by and rowwise do 按组计算累积和（cumsum） - Calculate cumulative sum (cumsum) by group Dplyr：group_by并将多列转换为向量 - Dplyr: group_by and convert multiple columns to a vector dplyr：group_by，对各个列求和，并基于分组行求和应用 function？ - dplyr: group_by, sum various columns, and apply a function based on grouped row sums?
相关标签
 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM