如何在 R 中使用 dcast 计算唯一计数答案

【问题标题】：how to calculate unique count using dcast in R如何在 R 中使用 dcast 计算唯一计数
【发布时间】：2020-11-15 10:22:25
【问题描述】：

我正在使用 dcast 转置下表

date               event          user_id
25-07-2020         Create          3455
25-07-2020         Visit           3567
25-07-2020         Visit           3567
25-07-2020         Add             3567
25-07-2020         Add             3678
25-07-2020         Add             3678
25-07-2020         Create          3567
24-07-2020         Edit            3871

我正在使用 dcast 转置以将我的事件作为列并计算 user_id

dae_summ <- dcast(ahoy_events, date ~ event, value.var="user_id")

但我没有获得唯一用户 ID。它多次计算相同的user_id。我该怎么做才能使同一日期和事件的一个 user_id 只计算一次。

【问题讨论】：

标签： r transpose dcast

【解决方案1】：

使用reshape 的基本 R 选项

out <- replace(
  u <- reshape(
    unique(transform(ahoy_events, user_id = ave(user_id, event, date, FUN = function(x) length(unique(x))))),
    direction = "wide",
    idvar = "date",
    timevar = "event"
  ),
  is.na(u),
  0
)

这样

> out
        date user_id.Create user_id.Visit user_id.Add user_id.Edit
1 25-07-2020              2             1           2            0
8 24-07-2020              0             0           0            1

数据

  "25-07-2020", "25-07-2020", "25-07-2020",
  "25-07-2020", "25-07-2020", "25-07-2020", "25-07-2020", "24-07-2020"
), event = c(
  "Create", "Visit", "Visit", "Add", "Add", "Add",
  "Create", "Edit"
), user_id = c(
  3455L, 3567L, 3567L, 3567L, 3678L,
  3678L, 3567L, 3871L
)), class = "data.frame", row.names = c(
  NA,
  -8L
))

【讨论】：

【解决方案2】：

使用reshape2 包，您可以使用以下内容：

library(reshape2)

数据：

zz <- "date               event          user_id
       25-07-2020         Create          3455
       25-07-2020         Visit           3567
       25-07-2020         Visit           3567
       25-07-2020         Add             3567
       25-07-2020         Add             3678
       25-07-2020         Add             3678
       25-07-2020         Create          3567
       24-07-2020         Edit            3871"
data <- read.table(text=zz, header = TRUE)

代码：

data %>% 
  dcast(user_id ~ event, value.var="user_id",fun.aggregate = function(x) length(unique(x)))

输出：

date         Add     Create      Edit      Visit
<fctr>       <int>   <int>       <int>     <int>
24-07-2020   0       0           1         0
25-07-2020   2       2           0         1

^{由reprex package (v0.3.0) 于 2020 年 7 月 25 日创建}

【讨论】：

【解决方案3】：

我们可以使用来自data.table的uniqueN

library(data.table)
dcast(setDT(ahoy_events), date ~ event, fun.aggregate = uniqueN)
#         date Add Create Edit Visit
#1: 24-07-2020   0      0    1     0
#2: 25-07-2020   2      2    0     1

或使用tidyr 中的pivot_wider 并将values_fn 指定为n_distinct

library(tidyr)
library(dplyr)
ahoy_events %>%
   pivot_wider(names_from = event, values_from = user_id, 
      values_fn = list(user_id = n_distinct), values_fill = list(user_id = 0))
# A tibble: 2 x 5
#   date       Create Visit   Add  Edit
#  <chr>       <int> <int> <int> <int>
#1 25-07-2020      2     1     2     0
#2 24-07-2020      0     0     0     1

数据

ahoy_events <- structure(list(date = c("25-07-2020", "25-07-2020", "25-07-2020", 
"25-07-2020", "25-07-2020", "25-07-2020", "25-07-2020", "24-07-2020"
), event = c("Create", "Visit", "Visit", "Add", "Add", "Add", 
"Create", "Edit"), user_id = c(3455L, 3567L, 3567L, 3567L, 3678L, 
3678L, 3567L, 3871L)), class = "data.frame", row.names = c(NA, 
-8L))

【讨论】：

【解决方案4】：

你可以试试：

library(reshape2)

#Data
df <- structure(list(date = c("25-07-2020", "25-07-2020", "25-07-2020", 
"25-07-2020", "25-07-2020", "25-07-2020", "25-07-2020", "24-07-2020"
), event = c("Create", "Visit", "Visit", "Add", "Add", "Add", 
"Create", "Edit"), user_id = c(3455L, 3567L, 3567L, 3567L, 3678L, 
3678L, 3567L, 3871L)), class = "data.frame", row.names = c(NA, 
-8L))

#New code
dae_summ <- dcast(df, date ~ event,  value.var="user_id",fun.aggregate = function(x) length(unique(x)))

        date Add Create Edit Visit
1 24-07-2020   0      0    1     0
2 25-07-2020   2      2    0     1

您的代码会产生这样的结果：

        date Add Create Edit Visit
1 24-07-2020   0      0    1     0
2 25-07-2020   3      2    0     2

所以还是有区别的。

【讨论】：