r-square intro

a few basic building blocks
Author

Wilfried Cools & Lara Stas

Published

January 3, 2024

SQUARE consultants
square.research.vub.be

Compiled on R 4.3.1

What-Why-Who

This site aims to introduce researchers to the tidyverse ecosystem in R.

Our target audience is primarily the research community of the VUB / UZ Brussel, particularly those who have some basic experience in R and want to know more.

We invite you to help improve this document by sending us feedback: square@vub.be

First Tidyverse Steps

  • Data manipulation and visualisation
    • simple but very important
    • bridges the gap between raw data and modeling
    • important part of most analyses
    • often neglected in statistics courses
  • Our focus: tidyverse
    • a set of R packages (~ functions)
    • in between raw data and modeling

tidyverse: Why it exists

  • R; a flexible open source statistical programming tool (2000)
    • open source: many contributors writing code their own way
    • users have to adapt to each package / function
  • Commit to shared rules (not reduce R flexibility)
    • consistency in terms of input and output
      • contract with user → tidy data
      • contract with developer
        • consistency & intuitive/sensible defaults
        • constancy of data type by default
    • consistency in function names and (order of) arguments
library(tidyverse)
  • tidyverse = first successful attempt to make R more consistent
    • earlier attempts failed
    • tidyverse well thought through
    • tidyverse makes sense for most
    • tidyverse supported and promoted by Rstudio
  • Ecosystem emerges, following the tidyverse rules
    • tibble for data representation
    • tidyr for tidying data
    • dplyr for manipulating data frames
    • ggplot for visualizing data
    • stringr for dealing with texts
    • readr for reading in data
    • forcats for dealing with factors
    • purrr for functional programming (advanced)
    • …

Find convenient cheat sheets here or directly in RStudio (Help → Cheat Sheets).

Set up tidyverse packages

  • The R-primer page (see menu: context) can maybe serve as a basic introduction on the use of R in general.

  • Install the tidyverse package (at least once)

install.packages('tidyverse')
  • Load the tidyverse package (once per R session)
    • the individual packages that are loaded by default are listed
    • conflicts are listed
library(tidyverse)
  • Conflicts result from identical function names
    • resolve conflicts
      • explicit referencing of package with ::
        • e.g., stat::filter( ) or dplyr::filter( )
      • creating new default
        • e.g., select <- dplyr::select
  • Conflicts can be checked for tidyverse
tidyverse_conflicts( )
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
  • tidyverse ecosystem includes
    broom, conflicted, cli, dbplyr, dplyr, dtplyr, forcats, ggplot2, googledrive, googlesheets4, haven, hms, httr, jsonlite, lubridate, magrittr, modelr, pillar, purrr, ragg, readr, readxl, reprex, rlang, rstudioapi, rvest, stringr, tibble, tidyr, xml2, tidyverse.
tidyverse_packages( )

tidy data (input)

  • Hadley Wickham’s ggplot (now works at RStudio)
    • consistent input, easier to write visualization functions
    • enforce use of ‘tidy’ data
  • Tidy data
    • observations in focus assigned a row, each
    • columns to add properties to these observations (cell values)
    • tabular, possibly disentangled into multiple tables
  • tibble
    • data.frame 2.0
    • do less, complain more

tidy output

  • Max Kuhn’s caret (now works at RStudio) turned into the broom package
    • homogenize statistical output
    • output to potentially serve as input
  • Tidy output
    • one-row model information glance
    • multiple row statistical summary tidy
    • model based extended data augment
  • Output also turned into a tibble

tibbles: the tidyverse data type

  • The tibble package offers the tidyverse data type, a tibble

  • A tibble is a data frame, not necessarily the other way around

  • A data frame is R’s data type for analysis

    • a list of equally sized vectors
      • numeric vector (either double, integer, or complex)
      • factor (ordered, not ordered)
      • boolean vector
      • character

  • A tibble enhances a data frame

    • for convenience and consistency
    • no row-names, must be part of data
    • different default behavior
      • printing, naming, …
      • less forgiving
    • example: print

  • Create tibble with tibble( ) or tribble( ) function

    • notice: class( ) shows both data.frame and tbl_df
    • notice: no row names, all info made explicit as data
    • compare with dataframe
mytibble <- tibble(
  colA = c("a","b","c"),
  colB = c(1:3)
)
(mytibble <- tribble(
  ~colA, ~colB,
  "a",   1,
  "b",   2,
  "c",   3
))
# A tibble: 3 × 2
  colA   colB
  <chr> <dbl>
1 a         1
2 b         2
3 c         3
class(mytibble)
[1] "tbl_df"     "tbl"        "data.frame"
mydf <- data.frame(colA=c('a','b','c'),colB=1:3)
class(mydf)
[1] "data.frame"
  • No need to think much about tibbles
    • a tibble is a data frame
    • tidyverse functions automatically enhance data frames to tibbles

pipes: a convenient way of chaining functions

  • The magrittr package offers the pipe function
    • %>% or |>
    • pushes left hand side into right hand side
      • eg., object %>% function
    • borrowed from functional programming
  • tidyverse always has as first argument it’s input
    • function(input, …)
    • pipes are convenient to chain functions
      • eg., object %>% function %>% function %>% function …
  • Pipes read from left to right
    • most base R use reads inside-out
    • compare
      • mtcars %>% pull(mpg) %>% mean()
      • mean(mtcars$mpg)
    • especially of interest with multiple steps, serves readability
    • example: root sum of squares for two sets of 10, sampled from standard normal
x1 <- rnorm(10); x2 <- rnorm(10)
sqrt(sum((x1-x2)^2))
[1] 4.570047
(x1-x2)^2 %>% sum( ) %>% sqrt( )
[1] 4.570047

Example: tidyverse

  • Create factors for all variables with fewer than 4 distinct values
    • for data.frame mtcars
    • change the elements (mutate)
      • for all variables (across)
        • where variable . < 4 distinct values
        • to factor
    • and show the structure (glimpse)
mtcars %>% 
    mutate(
        across(
            where(~n_distinct(.)<4),
            as.factor)) %>% 
    select(1:4) %>% glimpse
Rows: 32
Columns: 4
$ mpg  <dbl> 21.0, 21.0, 22.8, 21.4, 18.7, 18.1, 14.3, 24.4, 22.8, 19.2, 17.8,…
$ cyl  <fct> 6, 6, 4, 6, 8, 6, 8, 4, 4, 6, 6, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 8,…
$ disp <dbl> 160.0, 160.0, 108.0, 258.0, 360.0, 225.0, 360.0, 146.7, 140.8, 16…
$ hp   <dbl> 110, 110, 93, 110, 175, 105, 245, 62, 95, 123, 123, 180, 180, 180…