如何使用多个变量进行滚动总和（或滚动平均值）

Question

我是使用 zoo 包的新手，所以也许这是一个简单的问题。 我有以下数据框（df）：

library(lubridate)
library(zoo)
library(dplyr)

Date <- c("2010-01-28", "2010-01-28", "2010-02-28", 
           "2010-02-28", "2010-03-28", "2010-03-28",
           "2010-04-28", "2010-04-28")

Date <- as_date(Date)
Amount <- 1:8
Prod <- c("Corn", "Potato","Corn", "Potato","Corn", "Potato","Corn", "Potato")
df <- data.frame(Date, Prod, Amount)

print(df)

 Date         Prod    Amount
 2010-01-28   Corn      1
 2010-01-28 Potato      2
 2010-02-28   Corn      3
 2010-02-28 Potato      4
 2010-03-28   Corn      5
 2010-03-28 Potato      6
 2010-04-28   Corn      7
 2010-04-28 Potato      8

我想要的是计算每个变量的滚动总和，“窗口”为 3 天，然后制作一个新的数据框，如下所示：

 Date        Prod     Amount
 2010-03-28   Corn      9
 2010-03-28 Potato      12
 2010-04-28   Corn      15
 2010-04-28 Potato      18

可能rollapply() 和 dplyr 可以完成这项工作，但我不知道如何解决这个问题。 如果有人可以提供帮助，我将不胜感激:)

Answer 1

我是用dplyr::lag()做的

library(dplyr)
library(tibble)

## Data
data <- tribble(
  ~Date,        ~Prod,    ~Amount,
  "2010-01-28",   "Corn",      1,
  "2010-01-28", "Potato",      2,
  "2010-02-28",   "Corn",      3,
  "2010-02-28", "Potato",      4,
  "2010-03-28",   "Corn",      5,
  "2010-03-28", "Potato",      6,
  "2010-04-28",   "Corn",      7,
  "2010-04-28", "Potato",      8
)

# Code

data %>% 
  group_by(Prod) %>% 
  mutate(cum_amount = Amount + lag(Amount, 1) + lag(Amount, 2)) %>% 
  filter(!is.na(cum_amount))


# A tibble: 4 x 4
# Groups:   Prod [2]
  Date       Prod   Amount cum_amount
  <chr>      <chr>   <dbl>      <dbl>
1 2010-03-28 Corn        5          9
2 2010-03-28 Potato      6         12
3 2010-04-28 Corn        7         15
4 2010-04-28 Potato      8         18

为了您的评论更新

data %>% 
  group_by(Prod) %>% 
  mutate(cum_amount = c(rep(NA, 2), zoo::rollsum(Amount, 3))) %>% 
  filter(!is.na(cum_amount))

PS：记得在你的问题中包含 R 标签