基于不同列的R运行计数

Question

我想根据之前在 ColumnB 中出现的次数来获取 ColumnA 中某项的运行计数。理想情况下，这个计数也可以是 ColumnC 的子集。

例如，我想在这里获得获胜者之前的 LOSSES 或失败者之前的 WINS 的总和：

#create df
year <- c(2017, 2017, 2017, 2017, 2017, 2016, 2016, 2016, 2016, 2016)
winner <- c('sam', 'ryan', 'sally', 'sally', 'ryan', 'sally', 'mike', 'ryan', 'mike', 'sam')
loser <- c('mike', 'mike', 'ryan', 'sam', 'sam', 'mike', 'sally', 'mike', 'ryan', 'sally')
df <- data.frame(year, winner, loser)

#successul methods for getting winner's cumulative wins or loser's cumulative losses
df <- as.data.table(df)[, winner_wins := seq(.N), by = "winner"][]
df <- as.data.table(df)[, loser_losses := seq(.N), by = "loser"][]

#successul methods for getting winner's cumulative wins or loser's cumulative losses by year
df <- df %>% group_by(year, winner) %>% mutate(winner_wins = row_number())
df <- df %>% group_by(year, loser) %>% mutate(loser_losses = row_number())

#failed attempt to get winner's cumulative losses by year
df <- df %>% group_by(year) %>% mutate(winner_losses = cumsum(winner == loser & year == year))

我希望输出是我的原始数据框，但有四个新列：winner_cum_wins、winner_cum_losses、loser_cum_wins、loser_cum_losses。

Answer 1

这应该会为您提供所需的所有数据框：

library(tidyverse)
df %>% 
    group_by(year) %>% 
    mutate(match_id_year = row_number()) %>% 
    gather(outcome, name, -year, -match_id_year) %>% 
    arrange(year, match_id_year) %>% 
    group_by(year, name) %>% 
    mutate(cum_wins_year = cumsum(outcome == "winner"),
           cum_losses_year = cumsum(outcome == "loser"))

Answer 2

year <- c(2017, 2017, 2017, 2017, 2017, 2016, 2016, 2016, 2016, 2016)
winner <- c('sam', 'ryan', 'sally', 'sally', 'ryan', 'sally', 'mike', 'ryan', 'mike', 'sam')
loser <- c('mike', 'mike', 'ryan', 'sam', 'sam', 'mike', 'sally', 'mike', 'ryan', 'sally')
df <- data.frame(year, winner, loser)

#successul methods for getting winner's cumulative wins or loser's cumulative losses by year
df <- df %>% group_by(year, winner) %>% mutate(winner_wins = row_number())
df <- df %>% group_by(year, loser) %>% mutate(loser_losses = row_number())

我创建了以下函数来计算x 在y 中的先前出现次数。

count_wins_losses <- function(x,y){
  n = length(x)
  counts = numeric(n)
  for (i in 1:n){
    counts2 = numeric(i)
    for (j in 1:i){counts2[j] = sum(x[i] == y[j])}
    counts[i] = sum(counts2)
  }
  return(counts)
}

我使用split 将函数应用到每年。

# count the cummullative wins of the losers
loser_cum_wins <- df %>%
  split(year) %>%
  lapply(., function(x) count_winner_losses(x$loser, x$winner)) %>%
  unlist()

# count the cummulative losses of the winners
winner_cum_losses <- df %>%
  split(year) %>%
  lapply(., function(x) count_winner_losses(x$winner, x$loser)) %>%
  unlist()

arrangement 在这里完成，以便 df 和 loser_cum_wins 和 winner_cum_losses 中的年份匹配。

df <- arrange(df, year)
df$loser_cum_wins <- loser_cum_wins
df$winner_cum_losses <- winner_cum_losses
df

## A tibble: 10 x 7
## Groups:   year, loser [6]
#    year winner loser winner_wins loser_losses loser_cum_wins winner_cum_losses
#   <dbl> <chr>  <chr>       <int>        <int>          <dbl>             <dbl>
# 1 2016. sally  mike            1            1             0.                0.
# 2 2016. mike   sally           1            1             1.                1.
# 3 2016. ryan   mike            1            2             1.                0.
# 4 2016. mike   ryan            2            1             1.                2.
# 5 2016. sam    sally           1            2             1.                0.
# 6 2017. sam    mike            1            1             0.                0.
# 7 2017. ryan   mike            1            2             0.                0.
# 8 2017. sally  ryan            1            1             1.                0.
# 9 2017. sally  sam             2            1             1.                0.
#10 2017. ryan   sam             2            2             1.                1.

使用count_wins_losses() 函数的另一种方法是通过year 过滤df 并为每个拆分使用该函数，然后合并结果。

df2016 <- df %>%
  filter(year == 2016)
df2017 <- df %>%
  filter(year == 2017)

df2016$loser_cum_wins <- with(df2016, count_winner_losses(loser, winner))
df2016$winner_cum_losses <- with(df2016, count_winner_losses(winner, loser))
df2017$loser_cum_wins <- with(df2017, count_winner_losses(loser, winner))
df2017$winner_cum_losses <- with(df2017, count_winner_losses(winner, loser))
rbind(df2016,df2017)