数据清洗.转换为tidyverse

Question

我很好奇下面的代码是否可以转换为tidyverse代码。我试过了dplyr：：变异，但一直没能让它正常工作。

df$Gender[df$Gender == "M"] <- "Man"
df$Gender[df$Gender == "Male"] <- "Man"
df$Gender[df$Gender == "F"] <- "Woman"
df$Gender[df$Gender == "Female"] <- "Woman"
df$Gender[df$Gender == "M & F"] <- "Man and Woman"
df$Gender[df$Gender == "Male & Female"] <- "Man and Woman"

Answer 1

有一种方法，用dplyr::case_when()

df$Gender <- dplyr::case_when(
  df$Gender %in% c("M", "Male") ~ "Man", 
  df$Gender %in% c("F", "Female") ~ "Woman",
  df$Gender %in% c("M & F", "Male & Female") ~ "Man and Woman",
  TRUE ~ NA_character_)

或者，如果您想使用典型的dplyr::/magrittr::管道链方法：

df <- df %>% mutate(Gender = case_when(
  Gender %in% c("M", "Male") ~ "Man", 
  Gender %in% c("F", "Female") ~ "Woman",
  Gender %in% c("M & F", "Male & Female") ~ "Man and Woman",
  TRUE ~ NA_character_))

最后，一个提示:当有许多唯一的值需要分组时，使用case_when() (或嵌套的ifelse()，或子设置的赋值，等等)。可能会变得很乏味。避免这种痛苦的一种方法是使用命名向量来替换每个值，使用字典样式的“查找表”(非正式术语--参见关于“关联数组”的wiki中的一些背景)。根据我的经验，这通常是最干净的：

# the unique values 
gender_values <- c("M","Man","Male","F","Woman","Female","MF","male-female")

# associate unique values with our new labels: "m", "f", and "b"
gender_lkup <- setNames(c("m","m","m","f","f","f","b","b"), gender_values)

# suppose this is a column of a df 
raw_column <- sample(gender_values, 10, replace=TRUE)

# create a clean one with `gender_lkup` 
clean_column <- gender_lkup[raw_column]

# inspect the two vectors side-by-side
data.frame(original=raw_column, cleaned=clean_column)