Functions in R

What is a function?

A function allows you to efficiently save an operation you may have to use repeatedly, and to run it while only having to modify its inputs, or arguments.

I want to rescale the vectors a, b, c, and d so that the smallest value is 0 and the largest value is 1. Here's how I would do it in R:

(a-min(a, na.rm = TRUE)) / (max(a, na.rm = TRUE)-min(a, na.rm = TRUE))
(b-min(b, na.rm = TRUE)) / (max(b, na.rm = TRUE)-min(b, na.rm = TRUE))
(c-min(c, na.rm = TRUE)) / (max(c, na.rm = TRUE)-min(c, na.rm = TRUE))
(d-min(d, na.rm = TRUE)) / (max(d, na.rm = TRUE)-min(d, na.rm = TRUE))

This is not considered good coding; it requires a lot of copying/pasting, it increases the number of opportunities to make a typo mistake, and it is not efficient.

Functions in R

Solution: custom functions. Custom functions in R are written in the following manner:

name <- function(arguments) {
  body
}

In the case of our reshaping function, it would look like this:

# Example for the argument a:
# (a-min(a, na.rm = TRUE)) / (max(a, na.rm = TRUE)-min(a, na.rm = TRUE))
reshape_function <- function(x) {
  (x-min(x,na.rm = TRUE)) / (max(x,na.rm = TRUE)-min(x,na.rm = TRUE))
}
reshape_function(c(-10, 0, 10))

## [1] 0.0 0.5 1.0

reshape_function(c(1, 2, 3, 4, 5))

## [1] 0.00 0.25 0.50 0.75 1.00

Iterative Functions in R

What if I want to apply the same function to multiple objects?

What if I wanted to apply reshape_function() to 30 variables in my dataset? Rewriting it would be inefficient...

A possible, base R solution: use a "for" loop.

for(vector in list(c(-10, 0, 10), c(1, 2, 3, 4, 5))) {
    print(reshape_function(vector))
}

## [1] 0.0 0.5 1.0
## [1] 0.00 0.25 0.50 0.75 1.00

However, the use of "for" loops is discouraged in R because they have to run sequentially through each element of the object. To loop over large objects, this is much slower than alternative options, which directly apply to the whole object.

Iterative Functions in R

A more efficient solution: the function map() from the package purrr.

library(purrr)
purrr::map(
    .x = list(
        c(-10, 0, 10), c(1, 2, 3, 4, 5)
    ),
    .f = reshape_function
)

## [[1]]
## [1] 0.0 0.5 1.0
## 
## [[2]]
## [1] 0.00 0.25 0.50 0.75 1.00

Iterative Functions in R

Notice that the previous chunk's output is a list. If you want to output a different class of object, purrr has the functions map_chr(), map_dbl(), and map_df() amongst others.

map() has a shorthand to simplify its use: instead of writing the function out, you can replace function() {} with ~ and the function's argument with .x.

map_chr(
    c(1, 3),
    function(x) {
        x + 10
    }
)

## [1] "11.000000" "13.000000"

map_chr(
    c(1, 3),
    ~ .x + 10
)

## [1] "11.000000" "13.000000"

Iterative Functions in R

We can also apply map() to a data frame. If we do so, map() will apply the same function to every component of the data frame, i.e. its columns. In practice, this is equivalent to applying map() to a list of vectors.

library(palmerpenguins) # A fun practice dataset about penguins!
data(package = 'palmerpenguins')
# Let's take a glimpse at the dataset:
penguins %>% glimpse()

## Rows: 344
## Columns: 8
## $ species           <fct> Adelie, Adelie, Adelie, Adelie, Adelie, Adelie, Adel…
## $ island            <fct> Torgersen, Torgersen, Torgersen, Torgersen, Torgerse…
## $ bill_length_mm    <dbl> 39.1, 39.5, 40.3, NA, 36.7, 39.3, 38.9, 39.2, 34.1, …
## $ bill_depth_mm     <dbl> 18.7, 17.4, 18.0, NA, 19.3, 20.6, 17.8, 19.6, 18.1, …
## $ flipper_length_mm <int> 181, 186, 195, NA, 193, 190, 181, 195, 193, 190, 186…
## $ body_mass_g       <int> 3750, 3800, 3250, NA, 3450, 3650, 3625, 4675, 3475, …
## $ sex               <fct> male, female, female, NA, female, male, female, male…
## $ year              <int> 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007…

Iterative Functions in R

I want to reshape the numerical variables of the penguins dataset. Using map():

penguins_reshaped <- penguins %>%
  map_dfr(
    ~ if(is.numeric(.x)) {
      reshape_function(.x)
    } else {
      .x
    }
  )
penguins_reshaped %>% glimpse()

## Rows: 344
## Columns: 8
## $ species           <fct> Adelie, Adelie, Adelie, Adelie, Adelie, Adelie, Adel…
## $ island            <fct> Torgersen, Torgersen, Torgersen, Torgersen, Torgerse…
## $ bill_length_mm    <dbl> 0.25454545, 0.26909091, 0.29818182, NA, 0.16727273, …
## $ bill_depth_mm     <dbl> 0.6666667, 0.5119048, 0.5833333, NA, 0.7380952, 0.89…
## $ flipper_length_mm <dbl> 0.15254237, 0.23728814, 0.38983051, NA, 0.35593220, …
## $ body_mass_g       <dbl> 0.2916667, 0.3055556, 0.1527778, NA, 0.2083333, 0.26…
## $ sex               <fct> male, female, female, NA, female, male, female, male…
## $ year              <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…

Iterative Functions in R

The dplyr package offers us a function that makes it much easier to modify multiple dataset columns at once: across():

penguins_reshaped <- penguins %>%
  mutate(
    across(
      c(bill_length_mm, bill_depth_mm, flipper_length_mm, body_mass_g),
      ~ reshape_function(.x)
    )
  )
penguins_reshaped %>% glimpse()

## Rows: 344
## Columns: 8
## $ species           <fct> Adelie, Adelie, Adelie, Adelie, Adelie, Adelie, Adel…
## $ island            <fct> Torgersen, Torgersen, Torgersen, Torgersen, Torgerse…
## $ bill_length_mm    <dbl> 0.25454545, 0.26909091, 0.29818182, NA, 0.16727273, …
## $ bill_depth_mm     <dbl> 0.6666667, 0.5119048, 0.5833333, NA, 0.7380952, 0.89…
## $ flipper_length_mm <dbl> 0.15254237, 0.23728814, 0.38983051, NA, 0.35593220, …
## $ body_mass_g       <dbl> 0.2916667, 0.3055556, 0.1527778, NA, 0.2083333, 0.26…
## $ sex               <fct> male, female, female, NA, female, male, female, male…
## $ year              <int> 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007, 2007…

across() identifies columns by name, or can be combined with where() to identify columns by their class. The following are equivalent:

penguins %>%
  mutate(
    across(
      c(bill_length_mm, bill_depth_mm, flipper_length_mm, body_mass_g, year),
      ~ reshape_function(.x)
    )
  )
penguins %>%
  mutate(
    across(
      matches("length_mm|depth_mm|body_mass|year"),
      ~ reshape_function(.x)
    )
  )
penguins %>%
  mutate(
    across(
      where(is.numeric),
      ~ reshape_function(.x)
    )
  )

Iterative Functions in R

Beyond the scope of this class: sometimes we want to iterate functions that have more than one argument.

To do so, purrr provides the following functions: map2() (two arguments) and pmap() (unlimited arguments)

animals <- c("tiger", "elephant", "giraffe")
colors <- c("red", "orange", "blue")
map2( # paste0() concatenates its arguments into a single string
    animals, colors, ~ paste0("The ", .x, " is ", .y, ".")
)

## [[1]]
## [1] "The tiger is red."
## 
## [[2]]
## [1] "The elephant is orange."
## 
## [[3]]
## [1] "The giraffe is blue."

For more on iterative functions, I strongly recommend this blog post by Rebecca Barter.

A (Quick) Introduction to R Markdown

RStudio comes with a text and code compiler that allows you to craft narrated scripts, slide decks, blogs, books, and more: R Markdown.

Instead of creating an R Script, you can select "R Markdown" when creating a new document in RStudio.

R Markdown allows you to use the "markdown" markup language to format your text, while include code and outputs using code "chunks":

Background

"The survey, which started in 1981, seeks to use the most rigorous, high-quality research designs in each country. The WVS consists of nationally representative surveys conducted in almost 100 countries which contain almost 90 percent of the world’s population, using a common questionnaire. [...] WVS seeks to help scientists and policy makers understand changes in the beliefs, values and motivations of people throughout the world."

Survey Contents

End of Course Assignment (Due on Wednesday, April 30)

Links

Hadley Wickham, Mine Çetinkaya-Rundel & Garrett Grolemund, R for Data Science, 2e — Custom Functions

Hadley Wickham, “dplyr 1.0.0: working across columns”

Rebecca Barter, "Learn to purrr"

RStudio, RStudio Cheatsheets

Explore R Further

More Complex Data Manipulation

Iterative coding (using loops for repetitive code) – purrr ’s map function is your friend. I recommend Thomas Mock, “Functional programming in R with Purrr” to get you started.

User-made functions in R – At some point, it will become time-effective to create your own functions to apply to your work. Hadley Wickham, Advanced R Chapter 6 – Functions.

Publishing Your R Work

RStudio, “Introduction to RMarkdown”. Summarizes the uses and utility of the RMarkdown framework.

Yihui Xie, “xaringan Presentations” – book chapter and presentation. Introduction to xaringan , a package that allows you to create slide decks using R. Also explore the xaringanExtra package.

With RMarkdown, create books using bookdown or a blog using blogdown.

Explore R Further

Data Visualization Using Plots

The R community organizes “Tidy Tuesday”. This is a weekly challenge where users are provided a dataset and participants then swap graphs and scripts used to create their visualizations.

David Robinson’s Tidy Tuesday live screencasts on YouTube. The perfect resource to follow along and try to replicate a professional coder’s scripts.

Yan Holtz and Conor Healy, “From Data to Viz”. An amazing repository of methods to create different data visualizations using R.

Explore R Further

Geospatial Data Visualization

Robin Lovelace, Jakub Nowosad, and Jannes Muenchow, Geocomputation with R. A great introduction to manipulating geospatial data (shapefiles and rasters) in R.

Edzer Pebesma, “Simple Features for R”. An introduction to the sf package, commonly used for geospatial work in R.

Edzer Pebesma, “Plotting Simple Features”. How to use sf and ggplot2 to visualize data using maps.

Explore R Further

For those interested in conducting data work in the development world: Kristoffer Bjarkefur, Luiza Cardoso de Andrade, Benjamin Daniels, and Maria Ruth Jones, Development Research in Practice – The DIME Analytics Data Handbook. A comprehensive account of tools and instruments to conduct quantitative development research.

For those looking for more hands-on, real-world data work: Ben Baldwin, “A beginner’s guide to nflfastR”. How to download and explore NFL play-by-play data. This is how I learnt how to use R. Further tutorials using this data can be found at the “Open Source Football” blog.