map()系列函数与summarise_at()/mutate_at()的适用场景及相关问题咨询

Hey there! Let's break down your questions step by step—these are super common when getting comfortable with purrr and dplyr, so great call asking for clarity.

The key difference boils down to what kind of data you're working with:

• For regular vector columns in a data frame: Reach for summarise_at()/mutate_at() first. These dplyr functions are built specifically for batch operations on multiple vector columns. For example, if you want to standardize 3 numeric columns, you can write:

  df %>% mutate_at(vars(col1, col2, col3), ~scale(.))
  

  No need for map() here—these _at() functions handle the vector-wise iteration for you cleanly.
• For list columns in a data frame: You need map() (or its variants like map_dfr(), map_lgl()). List columns are columns where each cell is a list (e.g., a nested data frame, a model object, or a vector of variable length). summarise_at()/mutate_at() can't handle these because they expect vector inputs. For example, if you have a column of nested data frames and want to calculate the mean of a variable in each:

  df %>% mutate(mean_x = map_dbl(nested_col, ~mean(.$x, na.rm = TRUE)))
  

• Flexibility factor: map() is far more general—it works on standalone lists, vectors, or even single objects. The _at() functions are strictly for data frame columns and only work with vector types. Use map() when you need custom, row-wise or element-wise logic that the dplyr batch functions can't handle.

Absolutely not! nest() is just one way to create a list column (for grouped data), but map() has way more use cases:

• You can use it on a standalone list:

  map(list(1:5, 6:10), sum)
  

• You can use it to generate a list column from regular columns without nesting:

  df %>% mutate(parsed_data = map(json_strings, jsonlite::fromJSON))
  

• That said, nest() + map() is a super useful combo for grouped modeling/analysis (like fitting a model per group), but it's not a requirement for using map().

It's simpler than you think—just assign them directly or use mutate:

1. Create a list column from scratch:

   library(tibble)
   df <- tibble(
     id = 1:3,
     my_list = list(c(1, 2), c(3, 4, 5), c(6))
   )
   

2. Generate a list column from existing data:

   df <- df %>% mutate(split_groups = map(values, ~split(., . %% 2)))
   

3. Use nest() to create nested data frames (a type of list column):

   nested_df <- iris %>% nest(data = -Species)
   

   Here, the data column is a list where each element is a sub-data frame for one species.

The most common workflow uses nest() + map() for grouped data, but you can also work with standalone lists:

For grouped data in a data frame: #

1. Nest your data by the grouping variable:

   nested_mtcars <- mtcars %>% group_by(cyl) %>% nest()
   

2. Fit a model to each nested data frame with map():

   nested_mtcars <- nested_mtcars %>% mutate(model = map(data, ~lm(mpg ~ wt, data = .)))
   

3. Extract model results (using the broom package for tidy output):

   nested_mtcars <- nested_mtcars %>% 
     mutate(tidy_results = map(model, broom::tidy)) %>% 
     unnest(tidy_results)
   

For standalone lists of data frames: #

If you have a list of separate data frames, just pass the list directly to map():

list_of_dfs <- list(df1, df2, df3)
models <- map(list_of_dfs, ~lm(y ~ x, data = .))

No external links needed—here are some trusted, free resources from the R community:

• RStudio's Tidyverse Essentials: The purrr module walks through map() basics, list columns, and common workflows with clear examples.
• R-Ladies Online Talks: Search for talks on "purrr" or "nested data frames"—they often feature real-world use cases that make the concepts click.
• StatQuest with Josh Starmer: His purrr explainer uses animations to break down how map() works, making it easy to grasp even if you're new to functional programming.

内容的提问来源于stack exchange，提问作者CloverCeline