How to use this book

We begin with a short orientation in which we explain the structure of the book, the chapter template, and the two distinct kinds of chapter it contains. The template is the same as the one used in the sister volumes of the rgtlab curriculum series.

Two kinds of chapter

Chapters 1 through 10 are the course. They are written to be read one per week over a ten-week quarter, in order, and they assume nothing beyond one graduate methods course. Each carries roughly two hours of lecture material and four to six hours of assigned work. A reader who works through these ten chapters and their exercises has the working knowledge expected of a junior statistician joining a trial team, or of a clinical investigator who must read a protocol critically.

Chapters 11 through 28 are advanced topics. Each one takes a question that the course chapters raise but do not settle, and pursues it to a defensible answer using simulation evidence from a specific research compendium. Should a run-in period be used, and what does it cost in power? Where should the visits be placed in a longitudinal trial? Is site a covariate or a random effect? When does an exact test beat its asymptotic approximation? These chapters are self-contained and may be read in any order; each names its prerequisite course chapters at the top.

The division is deliberate. A first-year student reads the first ten. A second-year student, a doctoral student, or a practicing statistician reads the ones that bear on the trial in front of them.

Chapter template

Every content chapter follows the same nine-section structure:

Learning objectives. A bulleted list of the capabilities the reader should have after working through the chapter.
Orientation. Two to four paragraphs of prose framing, setting out what the chapter is, why it matters, and how it relates to adjacent chapters.
The statistician’s contribution. A front-loaded section articulating the judgments at the center of the chapter that no software and no language model can make on the reader’s behalf.
Content sections. The substantive material, broken into sections with descriptive headings. Collapsible Check your understanding callouts appear at natural pauses.
Worked example. A concrete trial, carried end to end through the chapter’s tools.
Collaborating with an LLM. Three prompt patterns paired with what to watch for and how to verify.
Principle in use. Two or three habits that define defensible work in this area.
Exercises. Five exercises ranging from short conceptual checks to extended applied work.
Further reading. Canonical references, modern applied treatments, and software documentation.

Advanced-topic chapters add a Provenance section naming the research compendium behind their simulation results.

Conventions

The visual cues used throughout the book are described on the Conventions page. Code blocks default to R and are display-only: they show what to run and what the run produced, but they are not re-executed when the book is rendered.

A note for the medical and epidemiology readers

Two habits will make the statistical chapters easier. The first is to read every formula as a sentence before reading it as algebra. The sample-size formula in Chapter 5 says that the number of patients needed goes up with the noise in the measurement and down with the size of the effect you are looking for, squared; the algebra only makes that sentence precise. The second is to keep asking what decision the number supports. A \(p\)-value supports a regulatory decision; a confidence interval supports a clinical one; a conditional power calculation supports a monitoring committee’s decision to continue or to stop. The book is organized around those decisions.

Companion repositories

The book repository at https://github.com/rgt47/clinical-trials contains the book source and its bibliography. The advanced-topic chapters cite research compendia maintained separately; each Provenance section names the compendium and the report within it that carries the full analysis.