Background

This is an unusual post for me, I have avoided writing about R Markdown because there are so many resources already available on the topic (e.g., here, here, and here). However, recently I ran a session on using RMarkdown for my colleagues in the Centre for Social Issues Research. The aim of this was to demonstrate the usefulness of R Markdown (and hopefully convert a few people). For this session I created a set of resources¹ aimed at making the transition from SPSS to R Markdown a bit easier. The statistics content of these resources is mainly just some of the simpler standard tests taught to psychology undergraduate students.

The complete resources are available on this project page on the OSF. The main purpose of the exercise was to provide people with the tools to create this pdf using this R Markdown template. My hope is that by using this template, SPSS users might make the tranistion to R, and R Markdown (with the help of the wonderful papaja package Aust (2017)).

R Markdown Basics

I started off the workshop by going through some of the basics of R Markdown. When working with R Markdown, there are three types of text to be concerned with.

1. Plain text: this is what you will use to write your manuscript. The formatting of this plain text is fairly straight forward (see this cheatsheet)²

2. Code chunks: these are denoted by three back ticks ``` at the top and the bottom of a chunk with commands in curly brackets in the title (on the same line as the firs set of back ticks). Code chunks allow you to run analyses without leaving your document. Output from code chunks can be included or omitted from your main output document.
3. In-line code: this is denoted by a single back tick to begin a piece of code and a single back tick to close it code goes here.

Each time you open a piece of code (a chunk or in-line code) you have to identify what language you want to code in. This is done by including the letter “r” in the curly brackets or immediately after the opening back tick.

Working with R

In order to show some of the basic functionality of R, I ran some analyses using the data sets that are built into R. This means that anyone should be able to reproduce the analyses conducted using the template I provided (without needing to worry about loading data from other files). I also provided another document with an accompanying template detailing the steps for inputting data into R. But this process is more prone to errors and if you are not familiar with R it can be a bit unintuitive.

##  [1] "mpg"  "cyl"  "disp" "hp"   "drat" "wt"   "qsec" "vs"   "am"   "gear"
## [11] "carb"

## [1] 20.09062

## [1] 6.026948

##       mean       sd  min  max len
## 1 20.09062 6.026948 10.4 33.9  32

## [1] 24.39231

Statistical tests

Now that we know some of the basics, we’ll try running some statistical tests.

Running a t-test

I’m going to use the mtcars dataset to illustrate how to run and write up a t-test, in such a way that:

1. there is no copy/paste error in reporting values; and
2. you can recycle code and text.

In addition to mpg (miles per gallon), the other variable I am interested in is am (transmission; 0 = automatic, 1 = manual).⁴

The question I’m going to look at is:

1. Is there a difference in miles per gallon depending on transmission?

## 
##  Welch Two Sample t-test
## 
## data:  x$mpg by x$am
## t = -3.7671, df = 18.332, p-value = 0.001374
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -11.280194  -3.209684
## sample estimates:
## mean in group 0 mean in group 1 
##        17.14737        24.39231

## [1] 0.001373638

##         t 
## -3.767123

##       df 
## 18.33225

Chi-square

To illustrate a chi-squared test I will test if there is an association between engine type vs (0 = V-shaped, 1 = straight) and transmission type (0 = automatic, 1 = manual). Again this is to illustrate the code/write up - not a real test.

• First create a table using the command table(x$am,x$vs)
• Running a chi-squared test using chisq.test(table(x$am,x$vs)) returns:

## 
##  Pearson's Chi-squared test with Yates' continuity correction
## 
## data:  table(x$am, x$vs)
## X-squared = 0.34754, df = 1, p-value = 0.5555

• As with the t-test, in order to report it using in-line code you need to save the test as an object e.g., c using c <- chisq.test(table(x$am,x$vs))
• Save the effect size using w <- sqrt(c$statistic/(length(x$mpg)*c$parameter))
• And calculate the observed power using pw <- pwr.chisq.test(w=w,N=length(x$mpg),df=(c$parameter),sig.level = .05)

ANOVA and Correlation

For details on the ANOVA check out the pdf using the R Markdown template on the OSF page.

Tables

Again using the mtcars dataset, I’m going to make a few tables. Bear in mind that the code for these tables is designed to be used along with the papaja package and rendered to pdf, so the html tables in this post will not be formatted as correctly as they should be (refer to the resources on the OSF to see what it should look like).

• Firstly save mtcars as an object using x <- mtcars
• We will make a table of engine type (V-shaped vs Straight) agains number of gears (3, 4, or 5) using table(x$vs,x$gear) which returns:

##    
##      3  4  5
##   0 12  2  4
##   1  3 10  1

• This table is useful to us, but it doesn’t really sit well in our document
• We can address this using the apa_table() function
• Before using this function we need to create a matrix that can be passed through the function using the command test <- as.data.frame.matrix(table(x$vs,x$gear))
• We can give names to the rows using test <-rownames<-(test, c("V-shaped","Straight"))
• And we can give names to the columns using test <-colnames<-(test, c("3 gears","4 gears","5 gears"))
• Finally we pass our object (matrix) test through the apa_table() function to render a table that will be embedded in our document
• The code below will generate the Table 1:

apa_table(
   test
   , align = c("l", "c", "c", "c")
   , caption = "Engine type and number of gears"
   , added_stub_head = "Engine Type"
   #, col_spanners = makespanners()
   
)

For more complex tables and example figures refer to the relevant section of the pdf and R Markdown template on the OSF page.

Using Citations

The citr can be used to cite while you write much like using endnote with Word. I have exported my Zotero Library (using BetterBibTex) to a .bib “My Library.bib” file in the directory I am working in. To cite I simply use @ and the citation key e.g., @haidt_emotional_2001 returns Haidt (2001). To enclose the entire citation in parenthesis use square brackets [@haidt_emotional_2001] gives: (Haidt 2001). The full reference is automatically added to the reference list (see below).

It is also generally good practice to cite R and the R packages you have used in your analyses. In the current post I used R (Version 3.6.1; R Core Team 2017) and the R-packages blogdown (Version 0.12; Xie, Hill, and Thomas 2017), bookdown (Version 0.10; Xie 2016), citr (Version 0.3.0; Aust 2016), desnum (Version 0.1.1; McHugh 2017), extrafont (Version 0.17; Chang 2014), ggplot2 (Version 3.2.0; Wickham 2009), knitr (Version 1.23; Xie 2015), lsr (Version 0.5; Navarro 2015), MASS (Version 7.3.51.4; Venables and Ripley 2002), papaja (Version 0.1.0.9842; Aust and Barth 2017), pwr (Version 1.2.2; Champely 2018), and scales (Version 1.0.0; Wickham 2016).