R - 循環多個變量 group_by

Question

我希望能夠遍歷 group_by 變量，以便按變量和總數的每個組合進行聚合，並將它們綁定為一個。

我在這里看到了類似的東西：
dplyr- 在 for 循環 r 中分組

我試圖通過幾種不同的方式修改上面的代碼，但我似乎無法讓它與交叉一起工作。

df <- data.frame(
  location = c(rep("UK", 5), rep("USA", 5)),
  industry = c(rep("RETAIL", 3), rep("TECH", 7)),
  department = c(rep("SALES", 4), rep("MANUFACTURING", 6)),
  pay = rnorm(10),
  tax = rnorm(10)
)

temp <- crossing(varA = c("location",""), varB = c("industry",""), varC = c("department",""))

data <- data.frame()
for(i in 1:nrow(temp)){
test <- df %>%
  group_by(!!temp[i,]) %>%
  summarise_at(c("pay", "tax"), sum, na.rm = TRUE)

data <- rbind.fill(test, data)
}

Answer 1

這是我認為您正在尋找的內容。 這是一個dplyr解決方案。

set.seed(10)
df <- data.frame(
  location = c(rep("UK", 5), rep("USA", 5)),
  industry = c(rep("RETAIL", 3), rep("TECH", 7)),
  department = c(rep("SALES", 4), rep("MANUFACTURING", 6)),
  pay = rnorm(10),
  tax = rnorm(10)
)

temp <- crossing(varA = c("location",""), varB = c("industry",""), varC = c("department",""))

data <- data.frame()
for(i in 1:nrow(temp)){
# extracts only non "" values from temp[i,] and unnames them (else group_by will use names)
  vars <- unname(unlist(temp[i,which(temp[i,] != "")]))
  test <- df %>%
    # tells tidyselect to use all columns that match the contents of vars
    group_by(across(all_of(vars))) %>% 
    summarise_at(c("pay", "tax"), sum, na.rm = TRUE)
  # union_all does what you want rbind.fill to do
  data <- union_all(test, data)
}
print(data, n = 20)
# A tibble: 20 x 5
# Groups:   location, industry [8]
   location industry department       pay   tax
   <chr>    <chr>    <chr>          <dbl> <dbl>
 1 UK       RETAIL   SALES         -1.54  1.62 
 2 UK       TECH     MANUFACTURING  0.295 0.741
 3 UK       TECH     SALES         -0.599 0.987
 4 USA      TECH     MANUFACTURING -3.07  0.348
 5 UK       RETAIL   NA            -1.54  1.62 
 6 UK       TECH     NA            -0.305 1.73 
 7 USA      TECH     NA            -3.07  0.348
 8 UK       NA       MANUFACTURING  0.295 0.741
 9 UK       NA       SALES         -2.14  2.61 
10 USA      NA       MANUFACTURING -3.07  0.348
11 UK       NA       NA            -1.84  3.35 
12 USA      NA       NA            -3.07  0.348
13 NA       RETAIL   SALES         -1.54  1.62 
14 NA       TECH     MANUFACTURING -2.77  1.09 
15 NA       TECH     SALES         -0.599 0.987
16 NA       RETAIL   NA            -1.54  1.62 
17 NA       TECH     NA            -3.37  2.08 
18 NA       NA       MANUFACTURING -2.77  1.09 
19 NA       NA       SALES         -2.14  2.61 
20 NA       NA       NA            -4.91  3.70

如果你想用某些東西替換你的NA值，說"ALL" ，你可以簡單地做到這一點：

data %>% replace(is.na(.), "ALL")
# A tibble: 20 x 5
# Groups:   location, industry [8]
   location industry department       pay   tax
   <chr>    <chr>    <chr>          <dbl> <dbl>
 1 UK       RETAIL   SALES         -1.54  1.62 
 2 UK       TECH     MANUFACTURING  0.295 0.741
 3 UK       TECH     SALES         -0.599 0.987
 4 USA      TECH     MANUFACTURING -3.07  0.348
 5 UK       RETAIL   ALL           -1.54  1.62 
 6 UK       TECH     ALL           -0.305 1.73 
 7 USA      TECH     ALL           -3.07  0.348
 8 UK       ALL      MANUFACTURING  0.295 0.741
 9 UK       ALL      SALES         -2.14  2.61 
10 USA      ALL      MANUFACTURING -3.07  0.348
11 UK       ALL      ALL           -1.84  3.35 
12 USA      ALL      ALL           -3.07  0.348
13 ALL      RETAIL   SALES         -1.54  1.62 
14 ALL      TECH     MANUFACTURING -2.77  1.09 
15 ALL      TECH     SALES         -0.599 0.987
16 ALL      RETAIL   ALL           -1.54  1.62 
17 ALL      TECH     ALL           -3.37  2.08 
18 ALL      ALL      MANUFACTURING -2.77  1.09 
19 ALL      ALL      SALES         -2.14  2.61 
20 ALL      ALL      ALL           -4.91  3.7

Answer 2

也許你正在尋找這個。 用循環試試這個tidyverse解決方案：

library(tidyverse)
#Data
df <- data.frame(
  location = c(rep("UK", 5), rep("USA", 5)),
  industry = c(rep("RETAIL", 3), rep("TECH", 7)),
  department = c(rep("SALES", 4), rep("MANUFACTURING", 6)),
  pay = rnorm(10),
  tax = rnorm(10)
)
#Loop
vars <- names(df)[1:3]
List <- list()
#Code df[,i]
for(i in 1:length(vars)){
  test <- df %>%
    group_by(eval(parse(text=vars[i]))) %>%
    summarise_at(c("pay", "tax"), sum, na.rm = TRUE)
  names(test)[1] <- 'var'
  #Var
  vardf <- data.frame(Mainvar=rep(vars[i],nrow(test)))
  test <- cbind(vardf,test)
  #Save
  List[[i]] <- test
  
}
#Bind all
mydf <- do.call(rbind,List)
rownames(mydf)<-NULL

輸出：

     Mainvar           var        pay       tax
1   location            UK -0.8347144 -1.719750
2   location           USA -2.8887471 -4.079747
3   industry        RETAIL  0.1327241 -1.212067
4   industry          TECH -3.8561856 -4.587430
5 department MANUFACTURING -4.5570133 -4.248031
6 department         SALES  0.8335518 -1.551466

Answer 3

vars <- names(df)[1:3]  
vars_subsets <- 0:length(vars) %>%
  map(~combn(vars, .x, simplify = FALSE)) %>%
  unlist(recursive = FALSE)

vars_subsets %>%
  map(~
        df %>% 
          {if(length(.x) > 0) group_by(., across(all_of(.x))) else .} %>% 
          summarise(pay = sum(pay, na.rm = TRUE), tax = sum(tax, na.rm = TRUE)) 
  ) %>%
  bind_rows() %>%
  select(all_of(vars), pay, tax)

給出：

> head(x)
location industry    department        pay        tax
1     <NA>     <NA>          <NA>  2.7641031  3.2347055
2       UK     <NA>          <NA> -0.2370619  3.5215502
3      USA     <NA>          <NA>  3.0011650 -0.2868447
4     <NA>   RETAIL          <NA>  1.3318324  0.4189127
5     <NA>     TECH          <NA>  1.4322707  2.8157928
6     <NA>     <NA> MANUFACTURING  2.8567654  0.7405478