dplyr 中 summarise() 函数的替代方案

Question

food_consumption %>% 
   group_by(food_category) %>% 
   summarise(mod= lm(co2_emmission ~ consumption))

运行此代码后，我收到以下错误

错误：列 mod 的长度必须为 1（汇总值），而不是 12

我怎样才能使它正确，并获得每个类别的回归结果？

Answer 1

我们可以将lm 模型的输出包装在list 中，因为它有很多组件，而summarise 想要返回每个组的长度为 1

library(dplyr)
food_consumption %>% 
     group_by(food_category) %>%
     summarise(mod= list(lm(co2_emmission ~ consumption)))

---

在dplyr的devel版本中，可以使用condense，它会自动返回一个list

food_consumption %>% 
     group_by(food_category) %>%
     condense(mod= lm(co2_emmission ~ consumption))

---

使用可重现的示例

mtcars %>% 
     group_by(cyl) %>%
     summarise(mod = list(lm(mpg ~ gear)))
# A tibble: 3 x 2
#    cyl mod   
#  <dbl> <list>
#1     4 <lm>  
#2     6 <lm>  
#3     8 <lm>  

或condense

mtcars %>% 
   group_by(cyl) %>%
   condense(mod = lm(mpg ~ gear))
# A tibble: 3 x 2
# Rowwise:  cyl
#    cyl mod   
#  <dbl> <list>
#1     4 <lm>  
#2     6 <lm>  
#3     8 <lm>  

为了得到系数

mtcars %>% 
   group_by(cyl) %>%
    condense(mod = lm(mpg ~ gear), Coef = coef(mod))
# A tibble: 3 x 3
# Rowwise:  cyl
#    cyl mod    Coef     
#  <dbl> <list> <list>   
#1     4 <lm>   <dbl [2]>
#2     6 <lm>   <dbl [2]>
#3     8 <lm>   <dbl [2]>

或者mutate和map

mtcars %>% 
    group_by(cyl) %>%
    summarise(mod = list(lm(mpg ~ gear))) %>% 
    mutate(Coef = map(mod, coef))
# A tibble: 3 x 3
#    cyl mod    Coef     
#  <dbl> <list> <list>   
#1     4 <lm>   <dbl [2]>
#2     6 <lm>   <dbl [2]>
#3     8 <lm>   <dbl [2]>

或者另一个选项是nest，然后是map，而不是list

library(purrr)
mtcars %>% 
    group_by(cyl) %>% 
    nest %>% 
    transmute(mod = map(data, ~ lm(mpg ~ gear, data = .x)))
# A tibble: 3 x 2
# Groups:   cyl [3]
#    cyl mod   
#  <dbl> <list>
#1     6 <lm>  
#2     4 <lm>  
#3     8 <lm>