Сумма предыдущих пяти лет

Question

Мне нужно агрегировать предыдущие 5 лет переменной N_C в каждой строке.

Например: 2017 год - Sum_Five_Years = 10 (2017) +21 (2015) +14 (2014) +16 (2013) = 61

Данные:

library(dplyr) 
DF<-data.frame(company = c("DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM","DEL MAR PHARM"),
           year= c("2017","2015","2015","2015","2013","2012","2012","2012","2010","2010","2015","2014","2014","2013","2013","2012"),
           N_C=   c("0","7","5","4","3","24","52","99","43","37","5","7","7","4","9","20"), Sum_Year = c("0","21","21","21","16","195","195","195","80","80","21","14","14","16","16","195"))
DF <- DF %>% arrange(year)

         company year N_C Sum_Year
1  DEL MAR PHARM 2010  43       80
2  DEL MAR PHARM 2010  37       80
3  DEL MAR PHARM 2012  24      195
4  DEL MAR PHARM 2012  52      195
5  DEL MAR PHARM 2012  99      195
6  DEL MAR PHARM 2012  20      195
7  DEL MAR PHARM 2013   3       16
8  DEL MAR PHARM 2013   4       16
9  DEL MAR PHARM 2013   9       16
10 DEL MAR PHARM 2014   7       14
11 DEL MAR PHARM 2014   7       14
12 DEL MAR PHARM 2015   7       21
13 DEL MAR PHARM 2015   5       21
14 DEL MAR PHARM 2015   4       21
15 DEL MAR PHARM 2015   5       21
16 DEL MAR PHARM 2017  10       10

Ожидаемый результат

DF$Sum_Five_Year <- cbind(c("80","80","275","275","275","275","291","291","291","305","305","246","246","246","246","61"))

> DF
         company year N_C Sum_Year Sum_Five_Year
1  DEL MAR PHARM 2010  43       80            80
2  DEL MAR PHARM 2010  37       80            80
3  DEL MAR PHARM 2012  24      195           275
4  DEL MAR PHARM 2012  52      195           275
5  DEL MAR PHARM 2012  99      195           275
6  DEL MAR PHARM 2012  20      195           275
7  DEL MAR PHARM 2013   3       16           291
8  DEL MAR PHARM 2013   4       16           291
9  DEL MAR PHARM 2013   9       16           291
10 DEL MAR PHARM 2014   7       14           305
11 DEL MAR PHARM 2014   7       14           305
12 DEL MAR PHARM 2015   7       21           246
13 DEL MAR PHARM 2015   5       21           246
14 DEL MAR PHARM 2015   4       21           246
15 DEL MAR PHARM 2015   5       21           246
16 DEL MAR PHARM 2017  10       10            61

Я пробовал следующий код, но он не работает:

 library(data.table)
 setDT(DF)  
 DF[, `:=` (Sum_Five_Year= sum(N_C)), by= list(company,cut(year, breaks = c(5), right = F))]

Любое предложение будет очень оценено :)

Answer 1

Без дополнительных пакетов вы можете использовать sapply.

В приведенном ниже коде предполагается, что Sum_Year уже создан.Вы можете применить следующее непосредственно к вашему примеру:

distinct(DF, company, year, Sum_Year) %>%
  group_by(company) %>%
  mutate(
    year = as.integer(as.character(year)),
    Sum_Five_Year = sapply(year, function(x) sum(Sum_Year[between(year, x - 5 + 1, x)]))
  ) %>%
  left_join(DF %>% select(-Sum_Year), by = c("company", "year"))

Вывод:

# A tibble: 16 x 5
# Groups:   company [?]
   company      year Sum_Year Sum_Five_Year   N_C
   <chr>       <int>    <int>         <int> <int>
 1 DELMARPHARM  2010       80            80    43
 2 DELMARPHARM  2010       80            80    37
 3 DELMARPHARM  2012      195           275    24
 4 DELMARPHARM  2012      195           275    52
 5 DELMARPHARM  2012      195           275    99
 6 DELMARPHARM  2012      195           275    20
 7 DELMARPHARM  2013       16           291     3
 8 DELMARPHARM  2013       16           291     4
 9 DELMARPHARM  2013       16           291     9
10 DELMARPHARM  2014       14           305     7
11 DELMARPHARM  2014       14           305     7
12 DELMARPHARM  2015       21           246     7
13 DELMARPHARM  2015       21           246     5
14 DELMARPHARM  2015       21           246     4
15 DELMARPHARM  2015       21           246     5
16 DELMARPHARM  2017       10            61    10

В противном случае вы можете сделать:

DF %>%
  group_by(company, year) %>%
  mutate(N_C = as.numeric(as.character(N_C))) %>%
  summarise(Sum_Year = sum(N_C)) %>%
  mutate(
    year = as.integer(as.character(year)),
    Sum_Five_Year = sapply(year, function(x) sum(Sum_Year[between(year, x - 5 + 1, x)]))
    ) %>%
  left_join(DF %>% select(-Sum_Year), by = c("company", "year"))

Если вы хотитеизбавиться от дублированного формата, просто пропустите объединение в конце:

DF %>%
  group_by(company, year) %>%
  mutate(N_C = as.numeric(as.character(N_C))) %>%
  summarise(Sum_Year = sum(N_C)) %>%
  mutate(
    year = as.integer(as.character(year)),
    Sum_Five_Year = sapply(year, function(x) sum(Sum_Year[between(year, x - 5 + 1, x)]))
    )

Вывод:

# A tibble: 6 x 4
# Groups:   company [1]
  company      year Sum_Year Sum_Five_Year
  <chr>       <int>    <dbl>         <dbl>
1 DELMARPHARM  2010       80            80
2 DELMARPHARM  2012      195           275
3 DELMARPHARM  2013       16           291
4 DELMARPHARM  2014       14           305
5 DELMARPHARM  2015       21           246
6 DELMARPHARM  2017       10            61