Or, Learning R: Why and How – A guide for absolute strangers

I am a retired electrical engineer (majoring in telecommunication) and have done a master’s in AI before pursuing a master’s in (primarily quantitative) psychology, which has led me to where I am, doing a Ph.D. in the methodology of psychological processes. I used to be quite proficient in Matlab (as I had enough of it in my EE and AI). However, after switching tracks to behavioral sciences, I now identify as a radical R evangelist.

Sadly, too many fellow psychologists, even those working in quantitative psychology, are total strangers to R—and that is genuinely heart-breaking.

When speaking to non-R users, one frequently hears only a handful of excuses why people are reluctant to start using R. Here, I try to address them. However, I will not talk much about R’s computational power and efficiency compared to other languages (more specifically, Matlab, Python, and Julia), as I believe an average user in psychology would not notice any issues regarding the speed or efficiency; as a pretty advanced user, I have not encountered any significant efficiency problems in the past few years. In what follows, I try to encourage you to learn the true and only way of doing statistics, which, as everyone knows, is R.

Disclaimer: only the first figure belongs to me; others can be removed upon request.

Why do you think I need to learn R?

SPSS and Matlab are products of filthy Capitalism: They are very costly—which you would probably not care if you use university/student licenses or pirate them—and they do not belong to the open-source culture. I believe they are barriers on our way to a utopian open science future. Moreover, in general, SPSS sucks big time; it is limited in its functionality and the analyses it can do. It shouldn’t be surprising that SPSS is getting closer and closer to its death, although the inertia

Matlab, on the other hand, is not well-suited for [quantitative] psychology; it lacks essential functions and packages and doesn’t handle the kind of data we encounter in social and behavioral sciences properly. Yes, it is a functional language for [hardcore] signal analysis and control engineering (and perhaps some numerical analyses/simulations in science and engineering). However, in many cases, Python and/or Julia are better alternatives. Also, although Matlab has always been a dominant language in engineering, there are indications that it is losing it to R.

One can spot these declines in Google Trends–although it is not a definite indicator of this claim. It is worth noting that Matlab owes much of its popularity to pirates, mostly users from countries (such as Iran) where there are no (proper) copyright laws in order. Just take a close look at the Google Trends link, and compare Iran’s rank for Matlab (1st) to its rank for R and Python (66th and 59th, respectively).

I am already familiar with Matlab/SPSS and I’m quite ‘happy’ with it.

Are you really happy, though?

People, especially SPSS users, are usually entrapped in an abusive relationship; they rely on it for their work and have to deal with its sufferings since they are too afraid to leave it and move on to a better alternative. Some are even suffering from some sort of Stockholm syndrome.

You are not alone there.

What can R offer that others do not?

R is fabulous. It is superior to other languages in countless ways when it comes to statistical analyses. Most importantly, it has a very active community and a ton of useful packages which satisfy most—if not all—of your needs. You can perform advanced statistical modelings and analyses that otherwise require special, costly software; you can easily do advanced latent trait modelings (IRT, EFA, CFA, and SEM) and make complex hierarchical generalized linear models. You also can go full Bayesian in R with JAGS. Except for some advanced dynamic structural equation models (e.g., with latent mean centering) that are better achieved by Mplus, you can do virtually anything with the power of R—you may even run (and automate) your Mplus analysis from within R.

With R (and its siblings, Rmarkdown and Shiny) you can do things so easily that is hard to imagine elsewhere. You can make excellent LaTeX or HTML reproducible (APA) papers, technical reports, books, and blogs with Rmarkdown, bookdown, and blogdown. You can create interactive apps with Shiny to collect experiment data, make tutorials, and make powerful BI dashboards, and deploy them to RStudio Shiny Server or your own server. You can build interactive maps and produce magnificent visualizations. You can do machine learning in R (this and this for a start) and you can, like Andrew Piper and others, use R in digital humanities.

But I’m too old for this :(

This graph (from Navarro’s intro slides) summarizes why you should still start learning R:

I can assure you that it is quite easy to reach the point where R surpasses the maximum reward you can get from SPSS, even if you start from an absolute zero. You do not need to learn all of the stuff mentioned above at once; you first lay a fundamental basis of R knowledge and then learn more advanced topics and techniques according to your needs.

Owkay, but I don’t know where to start :(

There are literally thousands of free resources to learn R out there; just have a look at Mine Dogucu’s learnR4free. This overload of information prevents most people from starting the journey. Hence, I only mention two handy (and relatively concise) resources which I have recently found:

1. Danielle Navarro’s R for Psychological Science, which is well-tailored for psychologists; and

2. Ben Whalley’s Just Enough R, which, as the name suggests, is just enough to kickstart your journey.

However, for an even faster, more efficient learning, you can also benefit from RStudio cheat sheets after you have the fundamental knowledge of R. I suggest to follow them in this order:

• Base R
• RStudio IDE Cheat Sheet
• Data Import Cheat Sheet
• Advanced R
• Data Transformation Cheat Sheet
• R Markdown Cheat Sheet
• Data Visualization Cheat Sheet
• Apply Functions Cheat Sheet

Now that you are beginning, I very strongly suggest mastering packages plyr and dplyr so you do not look like a troglodyte to other R users. The former offers a better way of writing apply functions. The latter is the best way of handling data intuitively and efficiently with pipe operator (%>%) and its data manipulation functions. You can gain a good understanding of plyr in Hadley Wickham’s paper in the Journal of Statistical Software (2011) or this tutorial. For dplyr, you may benefit from Garrett Grolemund’s data wrangling tutorial (webinar and slides).

You should also know that Google, Stack Overflow, and Stack Exchange are your allies. In +90% of the cases, you can find literally the exact pieces of code you need in less than a minute or two. If you couldn’t find a proper answer to your question (or had difficulty understanding what you have discovered), simply ask it on Cross Validated, where a kind stranger will help you quite soon. Interestingly, you can cite questions and answers you find on Stack Overflow and Stack Exchange; they give you BibTeX citation code. It’s worth mentioning that some people receive quicker responses if they post their questions on Reddit. In the end, it’s up to you.

But I’m super busy :( How long does it all take?

This is perhaps the most solid excuse to postpone learning useful skills, whether they are about programming (learning another language, or practicing a decent coding style), using a new tool (like Git versioning), or deepening your superficial theoretical background on a subject. One should never forget this lesson, which a King of Jerusalem, probably King Solomon, draws from his life experience:

If the ax is dull and its edge unsharpened, more strength is needed, but skill will bring success. – (Ecclesiastes, 10:10)

I think that the tutorials I mentioned above should take less than a couple days to finish. Also, you can always benefit from the cheat sheets when you need them. The data wrangling slides (dplyr) took me less than an hour (20-30 minutes?) to master, and plyr shouldn’t be much different.

To be amazed—and constantly surprized—by the power and wonders of R, I strongly suggest following R function a Day and R posts you might have missed! on Twitter (the latter has a monthly digest you can subscribe to).

Before you go,

As a gift for reading this far, I would like to introduce Ditto clipboard manager, a tool that helps you keep dozens (or even thousands) of copied/cut images and texts (also files!) in your clipboard with a very tiny overload on your PC. You can always search in the list, and you may use it as a draft pad for quick notes. It takes few minutes to figure it out, and you will soon wonder how you had managed to work with a computer without it.

Since Sharing is Caring, sharing this article is appreciated, and feel free to leave comments.