R中的复杂数据框计算

Question

我目前正在导入两个这样显示的表（以最基本的形式）

Table 1
State Month Account           Value
NY    Jan   Expected Sales    1.04
NY    Jan   Expected Expenses 1.02

Table 2
State Month Account    Value
NY    Jan   Sales      1,000
NY    Jan   Customers  500
NY    Jan   F Expenses 1,000
NY    Jan   V Expenses 100

我的最终目标是创建第三个数据框，其中包含前两行的值并根据函数计算第四列

NextYearExpenses = (t2 F Expenses + t2 V Expenses)* t1 Expected Expenses
NextYearSales = (t2 sales) * t1 Expected Sales

所以我想要的输出如下

State Month New Account Value
NY    Jan   Sales       1,040
NY    Jan   Expenses    1,122

我对 R 比较陌生，我认为 ifelse 语句可能是我最好的选择。我尝试过合并表格并使用简单的列函数进行计算，但没有真正的进展。

有什么建议吗？

Answer 1

你可能需要做一些数据整理，但没有什么特别的

require(dplyr)
Table1<-tibble(State=c("NY","NY"), Month=c("Jan","Jan"), Account=c("Expected Sales", "Expected Expenses"), Value=c(1.04,1.02))

Table2<-tibble(State=c("NY","NY","NY","NY"), Month=c("Jan","Jan","Jan","Jan"), Account=c("Sales", "Customers", "F Expenses","V Expenses"), Value=c(1000,500,1000,100))

我要做的第一件事是将帐户重命名为通用名称，即费用，这将有助于我稍后合并到 Table1

Table2$Account[Table2$Account=="F Expenses"]<-"Expenses"
Table2$Account[Table2$Account=="V Expenses"]<-"Expenses"

然后我使用 group_by 函数并按州、月和帐户分组并求和

Table2 <- Table2 %>% group_by(State, Month,Account) %>% 
summarise(Tot_Value=sum(Value)) %>% ungroup()
head(Table2)

## State Month Account   Tot_Value
##  <chr> <chr> <chr>         <dbl>
## 1 NY    Jan   Customers       500
## 2 NY    Jan   Expenses       1100
## 3 NY    Jan   Sales          1000

然后与表 1 中的帐户重命名类似

Table1$Account[Table1$Account=="Expected Sales"]<-"Sales"
Table1$Account[Table1$Account=="Expected Expenses"]<-"Expenses"

合并成第三个表，Table 3

Table3<- left_join(Table1,Table2)

使用 mutate 来做需要的操作

Table3 <- Table3 %>% mutate(Value2=Value*Tot_Value)
head(Table3)

## # A tibble: 2 x 6
##   State Month Account  Value Tot_Value Value2
##   <chr> <chr> <chr>    <dbl>     <dbl>  <dbl>
## 1 NY    Jan   Sales     1.04      1000   1040
## 2 NY    Jan   Expenses  1.02      1100   1122

Answer 2

这是我对dplyr 和tidyr 所做的。 首先，我将您的初始表格与rbind 组合成一个长格式表格。由于您对每个 Account 值都有唯一标识符，因此它们不需要是单独的表。接下来我group_byState 和 Month 对这些进行分组，假设最终您将拥有各种州/月份。接下来我summarise 基于您指定的 Account 值并创建了两个新列。最后为了把它变成你想要的长格式，我使用了tidyr 中的gather 从宽格式转换为长格式。您可以通过删除%>% 之后将这些命令分成更小的块，以便更好地了解每个步骤的作用。

library(dplyr)
library(tidyr)
rbind(df,df2) %>%
  group_by(State,Month) %>%
  summarise(Expenses = (Value[which(Account == "F Expenses")] + Value[which(Account == "V Expenses")]) * Value[which(Account == "Expected Expenses")],
            Sales = Value[which(Account == "Sales")] * Value[which(Account == "Expected Sales")]) %>%
  gather(New_Account,Value, c(Expenses,Sales))


# A tibble: 2 x 4
# Groups:   State [1]
#  State Month New_Account Value
#  <chr> <chr> <chr>       <dbl>
#1 NY    Jan   Expenses     1122
#2 NY    Jan   Sales        1040

Answer 3

我建议您查看the concept of "tidy data"，因为使用您目前拥有的结构处理数据存在一些真正的挑战。例如。创建 t3 应该只需要 2-3 行代码，所有这些只是为了解决您的数据架构：

library(tidyverse)

t1 <- data.frame(State = rep("NY", 2),
                 Month = rep(as.Date("2018-01-01"), 2),
                 Account = c("Expected Sales", "Expected Expenses"),
                 Value = c(1.04, 1.02),
                 stringsAsFactors = FALSE)

t2 <- data.frame(State = rep("NY", 4),
                 Month = rep(as.Date("2018-01-01"), 4),
                 Account = c("Sales", "Customers", "F Expenses", "V Expenses"),
                 Value = c(1000, 500, 1000, 100),
                 stringsAsFactors = FALSE)

t3 <- t2 %>% 
  spread(Account, Value) %>% 
  inner_join({
    t1 %>% 
      spread(Account, Value)
  }, by = c("State" = "State", "Month" = "Month")) %>% 
  mutate(NewExpenses = (`F Expenses` + `V Expenses`) * `Expected Expenses`,
         NewSales = Sales * `Expected Sales`) %>% 
  select(State, Month, Sales = NewSales, Expenses = NewExpenses) %>% 
  gather(Sales, Expenses, key = `New Account`, value = Value)