使用 R 中不同列的值透视数据

Question

抱歉，如果我重复发布，但我尝试了我在 stackoverflow 中看到的不同方法，但无法完全解决问题或理解我为什么会遇到它。

所以我有一个这样的数据集：

council_name <- c("Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barking and Dagenham","Barnet","Barnet")
period <- c("1st 2006", "1st 2006", "1st 2006", "1st 2006", "2nd 2006", "2nd 2006", "2nd 2006", "2nd 2006", "1st 2006", "1st 2006") 
category <- c ("glass", "fridges", "paper", "glass", "glass", "fridges", "paper", "glass", "glass", "fridges")
data <- c(333, 222, 100, 98, 450, 540, 33, 450, 560, 120)
category2 <- c ("collected", "collected", "collected", "no donors", "collected", "collected", "collected", "no donors", "collected", "collected")
df <- data.frame (council_name, period, category, category2, data)

而我想要的是这样的：

council_name <- c("Barking and Dagenham","Barking and Dagenham","Barnet")
period <- c("1st 2006", "2nd 2006", "1st 2006") 
glass <- c(333, 450, 560)
fridges <- c(222,540,120)
paper <- c(100, 33, NA)
no.donors <- c(98, 450, NA)
df.desired <- data.frame (council_name, period, glass, fridges, paper, no.donors)

我一直在尝试使用数据透视函数做多种事情，但事实上我需要从 category1 和 category2 中提取列名，但用 df 中同一列的值填充单元格，这给了我各种各样的问题。

非常感谢您的帮助！

Answer 1

以下方法是data.table的一种：

使用您的数据：

library(data.table)

setDT(df)

df[, sum(data), by = .(council_name, period, category, category2)][, dcast(.SD, council_name + period ~ category + category2, value.var = "V1")]

会产生什么

df
           council_name   period fridges_collected glass_collected glass_no donors paper_collected
1: Barking and Dagenham 1st 2006               222             333              98             100
2: Barking and Dagenham 2nd 2006               540             450             450              33
3:               Barnet 1st 2006               120             560              NA              NA

这基本上是您的数据，除了名称和列顺序的差异。这些可以通过 data.table 的 setnames 和 setcolorder 来修复。

里面有什么：

• sum(data), by = .(council_name... 将为理事会、时期、类别和类别2 的每个独特组合汇总您的数据。请考虑只运行这一位来查看输出：df[, sum(data), by = .(council_name, period, category, category2)]
• 第二部分，[, dcast(.SD, council_name + period ~ category + category2, value.var = V1)] 将之前生成的长表转换为宽表（您想要的格式）。它的意思是它希望行中的理事会名称和句点以及列中的类别和类别2，并且值变量是V1（V1是第一个链式命令中给出的虚拟名称，因为我刚刚写了sum(data)，没有给出它是一个名字）。

希望对你有帮助。

Answer 2

在这种情况下，您可以使用spread() 函数。代码如下：

spread(data=df, key=category, value=data)

Answer 3

这是一个tidyverse 解决方案，使用pivot_wider 将数据转换为宽格式，然后使用rename 更改列名。

library(tidyverse)

df %>%
  # Pivot from long to wide format using the first two columns as id cols and using both category and category2 columns to get the new column names
  pivot_wider(id_cols = c(council_name,period),
              names_from = c(category, category2),
              values_from = data) %>%
  # Rename the columns
  rename("glass" = "glass_collected",
         "fridges" = "fridges_collected",
         "paper" = "paper_collected",
         "no.donors" = "glass_no donors")

# A tibble: 3 x 6
# council_name         period   glass fridges paper no.donors
# <fct>                <fct>    <dbl>   <dbl> <dbl>     <dbl>
# 1 Barking and Dagenham 1st 2006   333     222   100        98
# 2 Barking and Dagenham 2nd 2006   450     540    33       450
# 3 Barnet               1st 2006   560     120    NA        NA

Answer 4

这是另一个在 dcast 内聚合的 data.table 方法：

library(data.table)
setDT(df)

dcast(df, council_name + period ~ category + category2, value.var = "data", fun.aggregate = sum)
#            council_name   period fridges_collected glass_collected glass_no donors paper_collected
# 1: Barking and Dagenham 1st 2006               222             333              98             100
# 2: Barking and Dagenham 2nd 2006               540             450             450              33
# 3:               Barnet 1st 2006               120             560               0               0