滚动平均值、1、2 和 3 滞后统计 - R
Rolling Average, 1, 2, and 3 Lag of Statistics - R
我的数据集图片如下：
对于从 HomeTeam 开始及以上的每个变量，我想要获得全时滞后滚动平均值、1 个记录滞后平均值、2 个记录滞后平均值以及每个团队的 3 个记录滞后平均值。下面是我在 Excel 中为 2016 年德州技术团队的 HomeTeam 变量所做的一个示例。
几个注意事项：
- 我将需要对右侧 HomeTeam 中的每个变量执行此操作。显然不需要它向左走。
- 这里有多年的数据，所以我们需要按年份和团队代码分组才能准确地做到这一点。
提前谢谢大家。
使用伪造的 + 廉价的 OCRed 数据（当 OP 给我们一个可重现的例子时我会很乐意更新）：
使用 tidyverse 进行争论和旋转，使用 zoo 进行滚动，尝试，
library(tidyverse)
library(zoo)
df_games <-
tibble(game_code = c(1:21),
team_code = c(1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3),
opponent_code = c(231, 306, 129, 27, 694, 277, 107, 1320, 440, 314, 465, 380, 164, 295, 528, 428, 458, 367, 736, 574, 772),
year = 2016,
date = seq.Date(from = as.Date("2016-01-01"), by = 7, length.out = 21),
team = c("TA", "TA", "TA", "TA", "TA", "TA", "TA", "TA", "TB", "TB", "TB", "TB", "TB", "TB", "TC", "TC", "TC", "TC", "TC", "TC", "TC"),
opp = c("FI", "I", "CM", "AS", "T", "Hi", "C", "P", "MS", "JS", "U", "M", "C", "I2", "OS", "M", "C2", "L", "V","R", "WK"),
home_team = c(0, 1, 0, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0),
points = c(34, 13, 3, 20, 13, 51, 31, 49, 20, 63, 13, 24, 21, 17, 30, 23, 70, 14, 13, 46, 14),
opp_points = c(13, 34, 49, 13, 20, 31, 51, 3, 17, 13, 63, 21, 24, 20, 23, 30, 14, 70, 10, 14, 46),
total = c(47, 47, 52, 33, 33, 82, 82, 52, 37, 76, 76, 45, 45, 37, 53, 53, 84, 84, 23, 60, 60),
mov = c(-21, 21, 46, -7, 7, -20, 20, -46, -3, -50, 50, -3, 3, 3, -7, 7, -56, 56, -3, -32, 32),
spread = c(-4, 4, 32, -22.5, 22.5, -22, 22, -32.1, -6.5, -40, 40, -27, 27, 65, -9.5, 9.5, -38.5, 38.5, 2.5, -16.5, 16.5),
book_total = c(61, 61, 52, 56, 56, 63, 63, 52, 37, 76, 76, 45, 45, 37, 57, 57, 59.5, 59.5, 46.5, 60, 60),
book_pts = c(32.5, 28.5, 10, 39.25, 16.75, 42.5, 20.5, 42, 21.75, 58, 18, 36, 9, 15.25, 33.25, 23.75, 49, 10.5, 22, 38.25, 21.75),
book_opp_pts = c(28.5, 32.5, 42, 16.75, 39.25, 20.5, 42.5, 10, 15.25, 18, 58, 9, 36, 21.75, 23.75, 33.25, 10.5, 49, 24.5, 21.75, 38.25),
rush_att = c(52, 21, 33, 43, 43, 35, 38, 38, 35, 44, 35, 50, 19, 38, 47, 34, 35, 31, 32, 30, 33),
rush_yard = c(246, 63, 44, 127, 184, 189, 248, 255, 225, 280, 108, 200, 40, 105, 150, 89, 272, 48, 121, 97, 105))
df_games %>%
pivot_longer(# Here we pivot from wide format into long format in order to efficiently create all the lagged values with a single call for all variables
cols = !c(game_code:home_team), names_to = "metric", values_to = "values") %>%
arrange(team_code, metric, date) %>%
group_by(team_code, year, metric) %>% #The next lines create the lagged means
mutate(cumean = lag(cummean(values)), # Here is the grouped cumulative mean up to that match
lag_01 = zoo::rollmean(x = values, k = 2, fill = NA, align = "right"),
lag_02 = zoo::rollmean(x = values, k = 3, fill = NA, align = "right"),
lag_03 = zoo::rollmean(x = values, k = 4, fill = NA, align = "right")) %>%
select(-values) %>%
pivot_wider(#Here we pivot back into wide format.
names_from = metric, values_from = c(lag_01:lag_03)) %>% arrange(team_code, date)
# A tibble: 21 x 48
# Groups: team_code, year [3]
game_code team_code opponent_code year date team opp home_team cumean_book_opp_pts cumean_book_pts
<int> <dbl> <dbl> <dbl> <date> <chr> <chr> <dbl> <dbl> <dbl>
1 1 1 231 2016 2016-01-01 TA FI 0 NA NA
2 2 1 306 2016 2016-01-08 TA I 1 28.5 32.5
3 3 1 129 2016 2016-01-15 TA CM 0 30.5 30.5
4 4 1 27 2016 2016-01-22 TA AS 1 34.3 23.7
5 5 1 694 2016 2016-01-29 TA T 0 29.9 27.6
6 6 1 277 2016 2016-02-05 TA Hi 0 31.8 25.4
7 7 1 107 2016 2016-02-12 TA C 0 29.9 28.2
8 8 1 1320 2016 2016-02-19 TA P 1 31.7 27.1
9 9 2 440 2016 2016-02-26 TB MS 1 NA NA
10 10 2 314 2016 2016-03-04 TB JS 1 15.2 21.8
# ... with 11 more rows, and 38 more variables: cumean_book_total <dbl>, cumean_mov <dbl>,
# cumean_opp_points <dbl>, cumean_points <dbl>, cumean_rush_att <dbl>, cumean_rush_yard <dbl>,
# cumean_spread <dbl>, cumean_total <dbl>, lag_01_book_opp_pts <dbl>, lag_01_book_pts <dbl>,
# lag_01_book_total <dbl>, lag_01_mov <dbl>, lag_01_opp_points <dbl>, lag_01_points <dbl>,
# lag_01_rush_att <dbl>, lag_01_rush_yard <dbl>, lag_01_spread <dbl>, lag_01_total <dbl>,
# lag_02_book_opp_pts <dbl>, lag_02_book_pts <dbl>, lag_02_book_total <dbl>, lag_02_mov <dbl>,
# lag_02_opp_points <dbl>, lag_02_points <dbl>, lag_02_rush_att <dbl>, lag_02_rush_yard <dbl>,
# lag_02_spread <dbl>, lag_02_total <dbl>, lag_03_book_opp_pts <dbl>, lag_03_book_pts <dbl>,
# lag_03_book_total <dbl>, lag_03_mov <dbl>, lag_03_opp_points <dbl>, lag_03_points <dbl>,
# lag_03_rush_att <dbl>, lag_03_rush_yard <dbl>, lag_03_spread <dbl>, lag_03_total <dbl>
我的数据集图片如下：
对于从 HomeTeam 开始及以上的每个变量，我想要获得全时滞后滚动平均值、1 个记录滞后平均值、2 个记录滞后平均值以及每个团队的 3 个记录滞后平均值。下面是我在 Excel 中为 2016 年德州技术团队的 HomeTeam 变量所做的一个示例。
几个注意事项：
- 我将需要对右侧 HomeTeam 中的每个变量执行此操作。显然不需要它向左走。
- 这里有多年的数据，所以我们需要按年份和团队代码分组才能准确地做到这一点。
提前谢谢大家。
使用伪造的 + 廉价的 OCRed 数据（当 OP 给我们一个可重现的例子时我会很乐意更新）：
使用 tidyverse 进行争论和旋转，使用 zoo 进行滚动，尝试，
library(tidyverse)
library(zoo)
df_games <-
tibble(game_code = c(1:21),
team_code = c(1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3),
opponent_code = c(231, 306, 129, 27, 694, 277, 107, 1320, 440, 314, 465, 380, 164, 295, 528, 428, 458, 367, 736, 574, 772),
year = 2016,
date = seq.Date(from = as.Date("2016-01-01"), by = 7, length.out = 21),
team = c("TA", "TA", "TA", "TA", "TA", "TA", "TA", "TA", "TB", "TB", "TB", "TB", "TB", "TB", "TC", "TC", "TC", "TC", "TC", "TC", "TC"),
opp = c("FI", "I", "CM", "AS", "T", "Hi", "C", "P", "MS", "JS", "U", "M", "C", "I2", "OS", "M", "C2", "L", "V","R", "WK"),
home_team = c(0, 1, 0, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0),
points = c(34, 13, 3, 20, 13, 51, 31, 49, 20, 63, 13, 24, 21, 17, 30, 23, 70, 14, 13, 46, 14),
opp_points = c(13, 34, 49, 13, 20, 31, 51, 3, 17, 13, 63, 21, 24, 20, 23, 30, 14, 70, 10, 14, 46),
total = c(47, 47, 52, 33, 33, 82, 82, 52, 37, 76, 76, 45, 45, 37, 53, 53, 84, 84, 23, 60, 60),
mov = c(-21, 21, 46, -7, 7, -20, 20, -46, -3, -50, 50, -3, 3, 3, -7, 7, -56, 56, -3, -32, 32),
spread = c(-4, 4, 32, -22.5, 22.5, -22, 22, -32.1, -6.5, -40, 40, -27, 27, 65, -9.5, 9.5, -38.5, 38.5, 2.5, -16.5, 16.5),
book_total = c(61, 61, 52, 56, 56, 63, 63, 52, 37, 76, 76, 45, 45, 37, 57, 57, 59.5, 59.5, 46.5, 60, 60),
book_pts = c(32.5, 28.5, 10, 39.25, 16.75, 42.5, 20.5, 42, 21.75, 58, 18, 36, 9, 15.25, 33.25, 23.75, 49, 10.5, 22, 38.25, 21.75),
book_opp_pts = c(28.5, 32.5, 42, 16.75, 39.25, 20.5, 42.5, 10, 15.25, 18, 58, 9, 36, 21.75, 23.75, 33.25, 10.5, 49, 24.5, 21.75, 38.25),
rush_att = c(52, 21, 33, 43, 43, 35, 38, 38, 35, 44, 35, 50, 19, 38, 47, 34, 35, 31, 32, 30, 33),
rush_yard = c(246, 63, 44, 127, 184, 189, 248, 255, 225, 280, 108, 200, 40, 105, 150, 89, 272, 48, 121, 97, 105))
df_games %>%
pivot_longer(# Here we pivot from wide format into long format in order to efficiently create all the lagged values with a single call for all variables
cols = !c(game_code:home_team), names_to = "metric", values_to = "values") %>%
arrange(team_code, metric, date) %>%
group_by(team_code, year, metric) %>% #The next lines create the lagged means
mutate(cumean = lag(cummean(values)), # Here is the grouped cumulative mean up to that match
lag_01 = zoo::rollmean(x = values, k = 2, fill = NA, align = "right"),
lag_02 = zoo::rollmean(x = values, k = 3, fill = NA, align = "right"),
lag_03 = zoo::rollmean(x = values, k = 4, fill = NA, align = "right")) %>%
select(-values) %>%
pivot_wider(#Here we pivot back into wide format.
names_from = metric, values_from = c(lag_01:lag_03)) %>% arrange(team_code, date)
# A tibble: 21 x 48
# Groups: team_code, year [3]
game_code team_code opponent_code year date team opp home_team cumean_book_opp_pts cumean_book_pts
<int> <dbl> <dbl> <dbl> <date> <chr> <chr> <dbl> <dbl> <dbl>
1 1 1 231 2016 2016-01-01 TA FI 0 NA NA
2 2 1 306 2016 2016-01-08 TA I 1 28.5 32.5
3 3 1 129 2016 2016-01-15 TA CM 0 30.5 30.5
4 4 1 27 2016 2016-01-22 TA AS 1 34.3 23.7
5 5 1 694 2016 2016-01-29 TA T 0 29.9 27.6
6 6 1 277 2016 2016-02-05 TA Hi 0 31.8 25.4
7 7 1 107 2016 2016-02-12 TA C 0 29.9 28.2
8 8 1 1320 2016 2016-02-19 TA P 1 31.7 27.1
9 9 2 440 2016 2016-02-26 TB MS 1 NA NA
10 10 2 314 2016 2016-03-04 TB JS 1 15.2 21.8
# ... with 11 more rows, and 38 more variables: cumean_book_total <dbl>, cumean_mov <dbl>,
# cumean_opp_points <dbl>, cumean_points <dbl>, cumean_rush_att <dbl>, cumean_rush_yard <dbl>,
# cumean_spread <dbl>, cumean_total <dbl>, lag_01_book_opp_pts <dbl>, lag_01_book_pts <dbl>,
# lag_01_book_total <dbl>, lag_01_mov <dbl>, lag_01_opp_points <dbl>, lag_01_points <dbl>,
# lag_01_rush_att <dbl>, lag_01_rush_yard <dbl>, lag_01_spread <dbl>, lag_01_total <dbl>,
# lag_02_book_opp_pts <dbl>, lag_02_book_pts <dbl>, lag_02_book_total <dbl>, lag_02_mov <dbl>,
# lag_02_opp_points <dbl>, lag_02_points <dbl>, lag_02_rush_att <dbl>, lag_02_rush_yard <dbl>,
# lag_02_spread <dbl>, lag_02_total <dbl>, lag_03_book_opp_pts <dbl>, lag_03_book_pts <dbl>,
# lag_03_book_total <dbl>, lag_03_mov <dbl>, lag_03_opp_points <dbl>, lag_03_points <dbl>,
# lag_03_rush_att <dbl>, lag_03_rush_yard <dbl>, lag_03_spread <dbl>, lag_03_total <dbl>