Appendix A — Getting started with R

This book is not an R tutorial. It is a book about Bayesian reasoning that happens to use R to do the arithmetic, and you can get most of the value from it without writing a line of code — see How to use the code in this book below.

But several readers of the draft told me the same thing: they wanted to follow along, and nothing in the book told them how to get to a point where they could. This appendix fixes that. It is a signpost, not a course — enough to get R running, plus pointers to the people who teach it properly.

A.1 Installing R and an editor

You need two things, and they are separate: R itself, which does the computing, and an editor, which is where you actually work. Installing only the first leaves you at a bare console; almost nobody works that way.

A.1.1 Step 1 — install R

Download it from https://cran.r-project.org. This is CRAN, the Comprehensive R Archive Network — the official home of R, run by the R Foundation. Pick your operating system, take the latest release, accept the defaults.

Make sure you are on the real site

R and its editors are free, which unfortunately makes them attractive bait for lookalike download sites. The two addresses worth trusting are cran.r-project.org for R and posit.co for the editor. If you arrived somewhere else via a search result, close the tab and type the address in directly.

If you work somewhere with a managed software catalogue, check it first — many organisations already offer R and RStudio there, sometimes as a supported enterprise build, which saves you the download and the conversation with IT.

A.1.2 Step 2 — install an editor

You have two good options, both free and both from Posit (the company that employs many of the tidyverse developers).

RStudio Desktop — https://posit.co/download/rstudio-desktop/. The standard choice, and the one nearly every tutorial you will find online assumes. Start here if you have no reason not to.
Positron — https://positron.posit.co. Posit’s newer editor, built on the same foundations as VS Code. It handles R and Python equally well and has a built-in AI assistant. It is what I wrote this book in.

Either works for everything in this book. If you are undecided, take RStudio: when you get stuck and search for help, the answer will be written for RStudio.

A.1.3 Step 3 — install the packages

Base R does relatively little on its own. Everything interesting comes from packages — bundles of functions somebody else wrote and shared. Run this once, in the console (the pane you type into directly), not in a code chunk:

install.packages(c("tidyverse", "peopleanalyticsdata", "skimr"))

That covers Chapters 1 to 3. The Bayesian modelling chapters need brms, which is a larger install because it compiles models behind the scenes:

install.packages("brms")

On some systems brms also needs a working C++ toolchain. If it complains, the current instructions are maintained at https://mc-stan.org/cmdstanr/articles/cmdstanr.html — follow those rather than anything you find on a forum, because this is one of the few areas where old advice goes stale quickly.

A.1.4 If you would rather not install anything

Posit Cloud (https://posit.cloud) runs RStudio in a browser with a free tier. Nothing to install, nothing to persuade IT to approve, and it is a perfectly reasonable way to work through the Foundations chapters. The free tier limits how many hours a month you get and how much memory a project can use, which is fine for Chapters 1 to 3 and will start to pinch once you are fitting Bayesian models.

A.2 The packages this book leans on

Two names appear constantly and are worth knowing properly, because the book treats them as background furniture.

A.2.1 tidyverse

The tidyverse is not one package but a collection of them, designed to work together and sharing a common style. When you run library(tidyverse) you load about eight at once. The ones you will see in this book:

Package	What it does
`dplyr`	Manipulating tables — filter rows, add columns, group and summarise. Most of the code in this book is `dplyr`.
`ggplot2`	Every chart in this book.
`tibble`	A modern data frame — the table type most functions here return.
`tidyr`	Reshaping tables between long and wide.
`readr`	Reading data files in.

The unifying idea is tidy data: one row per observation, one column per variable, one value per cell. Chapter 1 introduces this properly. The tidyverse is built around the assumption that your data is in that shape, which is why its functions compose so neatly with one another.

A.2.2 brms

brms — “Bayesian Regression Models using Stan” — is the package that fits nearly every model from Chapter 6 onwards. It lets you write a Bayesian model using the same formula syntax as R’s ordinary regression functions:

brm(sales ~ performance + (1 | region), data = salespeople)

and then does the genuinely hard part — translating that into Stan, compiling it, and sampling from the posterior — without you having to write Stan yourself. It is the reason this book can teach Bayesian modelling at all without a chapter of C++.

You do not need to understand any of that to read the book. Chapter 6 introduces brms where it is first needed, and Chapter 11 explains what the sampling underneath is actually doing.

A.3 How to use the code in this book

Readers arrive here from very different places, so here is the honest guidance.

New to R, or new to statistics. Run the code. Type it rather than pasting it, change a number, break it, see what the error says. The Foundations chapters are deliberately gentle for exactly this reason, and the fluency you build there is what makes the later chapters readable.
Comfortable with statistics but not with R. You can read past the code and still follow every argument in this book — the prose carries the reasoning and the output is always interpreted in words. What you should not skip is the Foundations part itself, because the Bayesian way of thinking about a probability is a genuine mental shift, and Chapters 2 and 3 are where that shift starts.
Comfortable with both. Skim Foundations for the framing, then start properly at Chapter 4.

On using an AI assistant to write the code

If you can describe your approach clearly in plain language, an AI assistant can usually turn it into working R. That is a legitimate way to work and I use it myself — but the description has to be yours. Knowing what you want to compute is the analytical skill; producing the syntax is not. Appendix B sets out how to do this well, and where it goes wrong.

A.4 Where to learn R properly

I am not going to teach R here when others do it better and keep it current. These are the resources I actually point people to.

A.4.1 The cheatsheets

Posit publishes one-page reference cards for each major package, and they are the single most useful thing on this list — print the dplyr one and keep it next to you.

https://posit.co/resources/cheatsheets/

The Data transformation with dplyr sheet is the one to start with. It covers filter(), select(), group_by(), summarise() and mutate() — including mutate(), which this book uses without ever teaching formally. mutate() adds or changes a column; it is the verb you reach for whenever an exercise asks you to derive something new from what you already have, such as bucketing a numeric variable into groups.

A.4.2 Books, free online

R for Data Science (Wickham, Çetinkaya-Rundel & Grolemund) — https://r4ds.hadley.nz. The standard introduction to the tidyverse, and the one to work through if you want to be able to write the code in this book yourself rather than read it.
Handbook of Regression Modeling in People Analytics (Keith McNulty) — https://peopleanalytics-regression-book.org. The frequentist companion to what this book does, on much of the same data, since the peopleanalyticsdata package is his.

A.4.3 Built-in help

Inside R, ?function_name opens the documentation for anything — ?mutate, ?rbinom, ?set.seed. The Examples section at the bottom is usually the fastest way in; the rest of the page is written for people who already know the answer.

Appendix C has the wider reading list, including the Bayesian texts this book draws on.

# Getting started with R {#sec-getting-started-with-r} This book is not an R tutorial. It is a book about Bayesian reasoning that happens to use R to do the arithmetic, and you can get most of the value from it without writing a line of code — see [How to use the code in this book](#sec-r-how-to-use) below. But several readers of the draft told me the same thing: they wanted to follow along, and nothing in the book told them how to get to a point where they could. This appendix fixes that. It is a signpost, not a course — enough to get R running, plus pointers to the people who teach it properly. ## Installing R and an editor {#sec-r-install} You need two things, and they are separate: **R** itself, which does the computing, and an **editor**, which is where you actually work. Installing only the first leaves you at a bare console; almost nobody works that way. ### Step 1 — install R Download it from **<https://cran.r-project.org>**. This is CRAN, the Comprehensive R Archive Network — the official home of R, run by the R Foundation. Pick your operating system, take the latest release, accept the defaults. ::: callout-tip ## Make sure you are on the real site R and its editors are free, which unfortunately makes them attractive bait for lookalike download sites. The two addresses worth trusting are `cran.r-project.org` for R and `posit.co` for the editor. If you arrived somewhere else via a search result, close the tab and type the address in directly. If you work somewhere with a managed software catalogue, check it first — many organisations already offer R and RStudio there, sometimes as a supported enterprise build, which saves you the download and the conversation with IT. ::: ### Step 2 — install an editor You have two good options, both free and both from Posit (the company that employs many of the tidyverse developers). - **RStudio Desktop** — <https://posit.co/download/rstudio-desktop/>. The standard choice, and the one nearly every tutorial you will find online assumes. Start here if you have no reason not to. - **Positron** — <https://positron.posit.co>. Posit's newer editor, built on the same foundations as VS Code. It handles R and Python equally well and has a built-in AI assistant. It is what I wrote this book in. Either works for everything in this book. If you are undecided, take RStudio: when you get stuck and search for help, the answer will be written for RStudio. ### Step 3 — install the packages Base R does relatively little on its own. Everything interesting comes from **packages** — bundles of functions somebody else wrote and shared. Run this **once**, in the console (the pane you type into directly), not in a code chunk: ```r install.packages(c("tidyverse", "peopleanalyticsdata", "skimr")) ``` That covers Chapters 1 to 3. The Bayesian modelling chapters need `brms`, which is a larger install because it compiles models behind the scenes: ```r install.packages("brms") ``` On some systems `brms` also needs a working C++ toolchain. If it complains, the current instructions are maintained at <https://mc-stan.org/cmdstanr/articles/cmdstanr.html> — follow those rather than anything you find on a forum, because this is one of the few areas where old advice goes stale quickly. ### If you would rather not install anything **Posit Cloud** (<https://posit.cloud>) runs RStudio in a browser with a free tier. Nothing to install, nothing to persuade IT to approve, and it is a perfectly reasonable way to work through the Foundations chapters. The free tier limits how many hours a month you get and how much memory a project can use, which is fine for Chapters 1 to 3 and will start to pinch once you are fitting Bayesian models. ## The packages this book leans on {#sec-r-packages} Two names appear constantly and are worth knowing properly, because the book treats them as background furniture. ### tidyverse The **tidyverse** is not one package but a collection of them, designed to work together and sharing a common style. When you run `library(tidyverse)` you load about eight at once. The ones you will see in this book: | Package | What it does | |---|---| | `dplyr` | Manipulating tables — filter rows, add columns, group and summarise. Most of the code in this book is `dplyr`. | | `ggplot2` | Every chart in this book. | | `tibble` | A modern data frame — the table type most functions here return. | | `tidyr` | Reshaping tables between long and wide. | | `readr` | Reading data files in. | The unifying idea is **tidy data**: one row per observation, one column per variable, one value per cell. Chapter 1 introduces this properly. The tidyverse is built around the assumption that your data is in that shape, which is why its functions compose so neatly with one another. ### brms **brms** — "Bayesian Regression Models using Stan" — is the package that fits nearly every model from Chapter 6 onwards. It lets you write a Bayesian model using the same formula syntax as R's ordinary regression functions: ```r brm(sales ~ performance + (1 | region), data = salespeople) ``` and then does the genuinely hard part — translating that into Stan, compiling it, and sampling from the posterior — without you having to write Stan yourself. It is the reason this book can teach Bayesian modelling at all without a chapter of C++. You do not need to understand any of that to read the book. Chapter 6 introduces `brms` where it is first needed, and Chapter 11 explains what the sampling underneath is actually doing. ## How to use the code in this book {#sec-r-how-to-use} Readers arrive here from very different places, so here is the honest guidance. - **New to R, or new to statistics.** Run the code. Type it rather than pasting it, change a number, break it, see what the error says. The Foundations chapters are deliberately gentle for exactly this reason, and the fluency you build there is what makes the later chapters readable. - **Comfortable with statistics but not with R.** You can read past the code and still follow every argument in this book — the prose carries the reasoning and the output is always interpreted in words. What you should not skip is the Foundations part itself, because the Bayesian way of thinking about a probability is a genuine mental shift, and Chapters 2 and 3 are where that shift starts. - **Comfortable with both.** Skim Foundations for the framing, then start properly at Chapter 4. ::: callout-note ## On using an AI assistant to write the code If you can describe your approach clearly in plain language, an AI assistant can usually turn it into working R. That is a legitimate way to work and I use it myself — but the description has to be *yours*. Knowing what you want to compute is the analytical skill; producing the syntax is not. @sec-llms sets out how to do this well, and where it goes wrong. ::: ## Where to learn R properly {#sec-r-learning} I am not going to teach R here when others do it better and keep it current. These are the resources I actually point people to. ### The cheatsheets Posit publishes one-page reference cards for each major package, and they are the single most useful thing on this list — print the `dplyr` one and keep it next to you. **<https://posit.co/resources/cheatsheets/>** The **Data transformation with dplyr** sheet is the one to start with. It covers `filter()`, `select()`, `group_by()`, `summarise()` and `mutate()` — including `mutate()`, which this book uses without ever teaching formally. `mutate()` adds or changes a column; it is the verb you reach for whenever an exercise asks you to derive something new from what you already have, such as bucketing a numeric variable into groups. ### Books, free online - **R for Data Science** (Wickham, Çetinkaya-Rundel & Grolemund) — <https://r4ds.hadley.nz>. The standard introduction to the tidyverse, and the one to work through if you want to be able to write the code in this book yourself rather than read it. - **Handbook of Regression Modeling in People Analytics** (Keith McNulty) — <https://peopleanalytics-regression-book.org>. The frequentist companion to what this book does, on much of the same data, since the `peopleanalyticsdata` package is his. ### Built-in help Inside R, `?function_name` opens the documentation for anything — `?mutate`, `?rbinom`, `?set.seed`. The **Examples** section at the bottom is usually the fastest way in; the rest of the page is written for people who already know the answer. @sec-further-reading has the wider reading list, including the Bayesian texts this book draws on.