我找到了
NHL shift data
并且想评估一个模型,在这个模型中,根据两个团队的具体情况,进球将遵循泊松分布。
我的观点是,我们已经很清楚谁能得分(进球和助攻),但也许有人真的很擅长帮助他的球队得分,而不去计分表上(也许是通过产生失误?)或者是非常擅长阻止对方得分。
我可以创建一个类似下面的“数据”的数据集。每支球队通常有5名队员,但我只放了2名让这个例子易于理解。
基本上,我每班都有一条线,我知道轮班的结果(目标),轮班持续时间,我有一个为球队(为球员)和对方球队(对球员)比赛的球员ID列表。
我想做什么
获取“数据”数据集,并创建“模型数据”,其中一个虚拟变量指示玩家是否在给定的轮班中处于冰上。然后,我将为我的泊松模型创建一个公式,该公式将包括所有的假人,并将其传递给模型。我也可以丢一个假人,一个假人,但我也可以让mgcv:gam为我做。
我怀疑这会涉及到一些!!和quos(),但我不知道该怎么做。
data <- tibble(
shift_id = c(1, 2, 3, 4, 5, 6, 7, 8,9,10),
shift_duration = c(12, 7, 30, 11, 14, 16, 19, 32,11,12),
goal_for = c(1, 1, 0, 0, 1, 1, 0, 0,0,0),
for_players = list(
c("A", "B"),
c("A", "C"),
c("B", "C"),
c("A", "C"),
c("B", "C"),
c("A", "B"),
c("B", "C"),
c("A", "B"),
c("B", "C"),
c("A", "B")
),
against_players = list(
c("X", "Z"),
c("Y", "Z"),
c("X", "Y"),
c("X", "Y"),
c("X", "Z"),
c("Y", "Z"),
c("X", "Y"),
c("Y", "Z"),
c("X", "Y"),
c("Y", "Z")
)
)
(black magic goes here)
model_data <- tibble(
shift_id = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10),
shift_duration = c(12, 7, 30, 11, 14, 16, 19, 32, 11, 12),
goal_for = c(1, 1, 0, 0, 1, 1, 0, 0, 0, 0),
for_player_A = c(1, 1, 0, 1, 0, 1, 0, 1, 0, 1),
for_player_B = c(1, 0, 1, 0, 1, 1, 1, 1, 1, 1),
for_player_C = c(0, 1, 1, 1, 1, 0, 1, 0, 1, 0),
against_player_X = c(1, 0, 1, 1, 1, 0, 1, 0, 1, 0),
against_player_Y = c(0, 1, 1, 1, 0, 1, 1, 1, 1, 1),
against_player_Z = c(1, 1, 0, 0, 1, 1, 0, 1, 0, 1)
)
mod.gam <- mgcv::gam(
data = model_data,
formula = goal_for ~ offset(log(shift_duration)) + for_player_A + for_player_B + for_player_C +
against_player_X + against_player_Y + against_player_Z,
family = poisson(link = log)
)
数据
如下所示:
> data
# A tibble: 10 x 5
shift_id shift_duration goal_for for_players against_players
<dbl> <dbl> <dbl> <list> <list>
1 1.00 12.0 1.00 <chr [2]> <chr [2]>
2 2.00 7.00 1.00 <chr [2]> <chr [2]>
3 3.00 30.0 0 <chr [2]> <chr [2]>
4 4.00 11.0 0 <chr [2]> <chr [2]>
5 5.00 14.0 1.00 <chr [2]> <chr [2]>
6 6.00 16.0 1.00 <chr [2]> <chr [2]>
7 7.00 19.0 0 <chr [2]> <chr [2]>
8 8.00 32.0 0 <chr [2]> <chr [2]>
9 9.00 11.0 0 <chr [2]> <chr [2]>
10 10.0 12.0 0 <chr [2]> <chr [2]>
模型数据
如下所示:
> model_data
# A tibble: 10 x 9
shift_id shift_duration goal_for for_player_A for_player_B for_player_C against_player_X against_player_Y against_player_Z
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 1.00 12.0 1.00 1.00 1.00 0 1.00 0 1.00
2 2.00 7.00 1.00 1.00 0 1.00 0 1.00 1.00
3 3.00 30.0 0 0 1.00 1.00 1.00 1.00 0
4 4.00 11.0 0 1.00 0 1.00 1.00 1.00 0
5 5.00 14.0 1.00 0 1.00 1.00 1.00 0 1.00
6 6.00 16.0 1.00 1.00 1.00 0 0 1.00 1.00
7 7.00 19.0 0 0 1.00 1.00 1.00 1.00 0
8 8.00 32.0 0 1.00 1.00 0 0 1.00 1.00
9 9.00 11.0 0 0 1.00 1.00 1.00 1.00 0
10 10.0 12.0 0 1.00 1.00 0 0 1.00 1.00
模型结果:
Family: poisson
Link function: log
Formula:
goal_for ~ offset(log(shift_duration)) + for_player_A + for_player_B +
for_player_C + against_player_X + against_player_Y + against_player_Z
Parametric coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -22.0296 4317.9341 -0.005 0.996
for_player_A 0.0000 0.0000 NA NA
for_player_B -2.3026 2.0000 -1.151 0.250
for_player_C -0.1542 1.4142 -0.109 0.913
against_player_X 1.6094 1.4142 1.138 0.255
against_player_Y 0.0000 0.0000 NA NA
against_player_Z 20.2378 4317.9339 0.005 0.996
Rank: 5/7
R-sq.(adj) = 0.353 Deviance explained = 73.6%
UBRE = 0.26435 Scale est. = 1 n = 10
