Plot columns repeatedly

/ 2 mins read
2018-11-02

Suppose you have a large number of columns of a dataframe, and you want to plot each column – say a histogram for each column.

This post shows some ways of achieving this.

Let’s take the mtcars dataset as an example.

data(mtcars)

We will use the tidyverse approach:

library(tidyverse)

Way 1

mtcars %>% 
  select_if(is_numeric) %>% 
  map2(., names(.), ~ {ggplot(data = data_frame(.x), 
                              aes(x = .x)) + 
      geom_histogram() +
      labs(x= .y)})
#> $mpg
#> 
#> $cyl
#> 
#> $disp
#> 
#> $hp
#> 
#> $drat
#> 
#> $wt
#> 
#> $qsec
#> 
#> $vs
#> 
#> $am
#> 
#> $gear
#> 
#> $carb

Some explanations:

  • First, we take the dataset mtcars.
  • Then, we map a function (ie., ggplot()) to each column of mtcars, but we also parse the names of mtcars.
  • ggplot() likes dataframes, but map() serves lists/vectors, so we have to enshrine each vector to a dataframe using data_frame().
  • The data comes from the first list (mtcars), that’s where .x comes from (or refers to).
  • The names come from the second lsit (names(mtcars)), that’s where .y points to.

Way 2

A maybe more simple is this:

mtcars %>% 
  gather(key = item, value = value) %>% 
  ggplot() +
  aes(x = value) +
  geom_density() +
  facet_wrap(~ item, ncol = 2, scales = "free")

Test if column is normally distributed before doing anything else

Suppose we want to check whether a column is nicely normally distributed before plotting. That’s one way to checking that:

mtcars %>% 
  map(~ shapiro.test(.x)) %>% 
  map("p.value") %>% 
  keep(. > .05)
#> $mpg
#> [1] 0.1228814
#> 
#> $drat
#> [1] 0.1100608
#> 
#> $wt
#> [1] 0.09265499
#> 
#> $qsec
#> [1] 0.5935176