使用 dplyr 以数值变量为条件改变因子变量

Question

我有一个运动队的数据，用于跟踪与其他团队的胜负，结构如下：

Game TotalWins TotalLosses Team1Win Team1Loss Team2Win Team2Loss ...
1    1         0           1        NA        NA       NA
2    1         1           NA       NA        NA       1
3    2         1           NA       NA        1        NA
4    2         2           NA       1         NA       NA
5    3         2           NA       NA        1        NA
...

我想创建一个因子变量，其中包括比赛对阵的球队，以便数据看起来像这样：

Game TotalWins TotalLosses Team1Win Team1Loss Team2Win Team2Loss Team
1    1         0           1        NA        NA       NA        Team1
2    1         1           NA       NA        NA       1         Team2
3    2         1           NA       NA        1        NA        Team2
4    2         2           NA       1         NA       NA        Team1
5    3         2           NA       NA        1        NA        Team2
...

我的想法（不可行的代码）本质上是这样的：

if (Team1Win == 1 | Team1Loss == 1), Team = "Team1"
if (Team2Win == 1 | Team2Loss == 1), Team = "Team2"

我真的很纠结如何在 dplyr 中使用 mutate 来做到这一点。我用 ifelse、recode 等尝试了各种方法，但要么不断收到错误，要么得到我不想要的结果。

在 dplyr 中进行这项工作的正确和最有效的方法是什么？

Answer 1

与其他答案类似，但有一些有用的变化：

• na.rm = TRUE 在gather
• sub 工作得很好，不需要stringr
• 根据目标包含完整数据，使用full_join。

library(dplyr)
library(tidyr)

df = read.delim(text = 
"Game TotalWins TotalLosses Team1Win Team1Loss Team2Win Team2Loss
1 1 0 1 NA NA NA
2 1 1 NA NA NA 1
3 2 1 NA NA 1 NA
4 2 2 NA 1 NA NA
5 3 2 NA NA 1 NA", sep = " ")

df %>% 
  select(-starts_with("Total")) %>% 
  gather(Team, one, -Game, na.rm = TRUE) %>% 
  select(-one) %>% 
  mutate(Team = sub("Win|Loss", "", Team)) %>% 
  full_join(df, .)

#> Joining, by = "Game"
#>   Game TotalWins TotalLosses Team1Win Team1Loss Team2Win Team2Loss  Team
#> 1    1         1           0        1        NA       NA        NA Team1
#> 2    2         1           1       NA        NA       NA         1 Team2
#> 3    3         2           1       NA        NA        1        NA Team2
#> 4    4         2           2       NA         1       NA        NA Team1
#> 5    5         3           2       NA        NA        1        NA Team2

Answer 2

目前，我对 dplyr 的处理方法很感兴趣，因此我提供了一个使用 dplyr 的解决方案，可以扩展到您可能拥有的许多团队。它还使用了tidyr 和stringr，正如下面apom 在评论中有用地指出的那样。

library(dplyr)
library(tidyr)
library(stringr)

df = read_delim(
"Game TotalWins TotalLosses Team1Win Team1Loss Team2Win Team2Loss
1 1 0 1 NA NA NA
2 1 1 NA NA NA 1
3 2 1 NA NA 1 NA
4 2 2 NA 1 NA NA
5 3 2 NA NA 1 NA",delim = " ")

df %>% 
  gather("Team",value,contains("Team")) %>% 
  filter(!is.na(value)) %>%
  mutate(Team = str_replace_all(Team,c("Win" = "","Loss" = ""))) %>%
  select(-value)

Answer 3

你可以用一个简单的循环来做到这一点：

x = colnames(df)
df$team<- NA

for (i in 1:nrow(df))
{
 df$team[i] = x[which(df[i,]==1)]
}

最后你可以使用以下函数修剪“赢”和“输”：

df$team<- gsub("win", "",df$team)
df$team<- gsub("loss", "",df$team)

Answer 4

我很确定您的数据中有两个以上的团队，并且团队名称不是通用的。您要做的是首先将数据重塑为长格式并提取相关的团队名称。因此，您可能希望按照以下方式进行操作。

library(dplyr)
library(tidyr)
new_df <- df %>% 
  gather(team,idx,Team1Win:Team100Loss) %>%
  filter(!is.na(idx)) %>%
  select(-idx) %>%
  mutate(team = gsub("Win|Loss","",team))

如果你想保留那些宽列，那么你可以将新的 DF 加入旧的。

Answer 5

这可能是您正在寻找的。 （并非仅针对 2 个团队进行硬编码）

# solution 1 :
paste0("Team",ceiling(apply(df[-c(1:3)], 1, function(x) which(!is.na(x)))/2))
[1] "Team1" "Team2" "Team2" "Team1" "Team2"

# solution 2: using a apply() {basically a for loop itself]
apply(df[-c(1:3)], 1, function(x) gsub("(Team\\d)+.*", "\\1", colnames(df[-c(1:3)])[which(!is.na(x))]))
# [1] "Team1" "Team2" "Team2" "Team1" "Team2"

# solution 3: (long route to dplyr) [ you have indirectly taught me a lot in dplyr through my search for this solution]
func <- function(x){
  y = which(x == 1)             # get the location of where 1 appears
  z = rep(0, times = length(x)) # create a vector of 0's+location of 1
  z[y] = y                      # i.e. c(0,0,3,0,5) for Team2Win 
  z
}

df1 = df[-c(1:3)] %>% gather("key", "value", starts_with("Team")) %>% 
                group_by(key) %>%
                dplyr::mutate(x = func(value)) %>%
                filter(x != 0) %>% arrange(x) %>% select(key)

df$newcol = gsub("(Team\\d+).*", "\\1", df1$key)