Preface

Why this book?

Most biostatistics graduate curricula teach statistical theory thoroughly and practical data-analysis craft casually. Students learn likelihood theory and the fine points of hypothesis testing but arrive at their first collaborative project unsure how to set up a project directory, commit changes with git, parameterize a report, or write a defensible statistical analysis plan.

The omission is not an oversight by any individual program. It is structural. A curriculum committee allocates credit hours to what can be examined, and the craft of analysis has resisted examination: there is no closed-form answer to ‘how should this project be organized’, and no problem set that distinguishes a compendium a stranger can rebuild from a directory of scripts that happens to have produced the right numbers once. The result is that the skills a biostatistician uses every working day are the ones the degree least prepared them for, and they are acquired, if at all, by apprenticeship and accident.

This book aims to close that gap. It teaches the practical skills needed to do reproducible biostatistical work in 2026: the tools, the workflows, and the professional norms. We assume throughout that the reader will also be studying a theory-heavy statistical computing course alongside this one, and we treat modeling as that course’s territory rather than our own.

Why now

A book of this kind would have been useful in 2015. It is necessary in 2026, because three developments have converged that together change what a competent biostatistician is expected to be able to do.

Reproducibility stopped being a virtue and became a requirement. For two decades the case for reproducible analysis was made on scientific grounds and heard as exhortation (Goodman et al., 2016; Peng, 2011). That changed when the funders wrote it down. The 2022 memorandum from the White House Office of Science and Technology Policy (Office of Science and Technology Policy, 2022) directs every federal research funder to require immediate public access to publications and to the data underlying them, with agency implementation due by the end of 2025. Since January 2023 every NIH application has carried a Data Management and Sharing Plan that becomes a term of the award. The FDA has required CDISC-conformant submissions since 2016. A graduate entering the field now will be asked, in their first year, to name a repository, produce a persistent identifier, and state how the computational environment will be preserved. These are not questions a theory sequence answers, and the person on the team most likely to be asked them is the statistician. 33  Federal Requirements and Deposition treats the policy landscape directly; the rest of the book is, in effect, the technical answer to it.

Large language models moved the scarce skill. By 2026 a competent model will produce working R for most of the exercises in this book, and for a substantial share of the analysis code a junior statistician is asked to write. The bottleneck has moved from producing code to auditing it. This is a genuine change in what expertise consists of, and it cuts in a specific direction: the parts of the work a model does well are the mechanical parts this book teaches quickly, and the parts it does badly are the judgments this book spends most of its pages on. A model will not tell you that the verb in your discussion section outruns your design, that the join you just wrote inflated the denominator, or that the subgroup you are about to report was not one you were powered for. It will, however, write the join. Pretending the tools do not exist would leave students to discover their failure modes in production; teaching them without teaching verification would be worse. 28  AI-Assisted Coding takes the question head on, and every chapter closes with prompts paired with an explicit verification step.

The tooling finally settled. The infrastructure this book teaches was, until recently, a moving target. Quarto reached maturity and absorbed R Markdown’s role; renv (Ushey & Wickham, 2026) displaced packrat and became the default for project-level dependency pinning; the rocker images (Boettiger & Eddelbuettel, 2017) made containerized R routine rather than exotic; targets (Landau, 2021) gave R a dependency-aware build tool that fits how analysts actually work. A book written in 2018 would have taught tools that are now deprecated. The stack described here is stable enough to be worth learning and recent enough that most working biostatisticians have not been taught it.

What the peer curricula do and do not cover

Several chapters of this book justify their existence by reference to a survey of the published graduate curricula of peer United States biostatistics programs. The sources are recorded in Appendix B — The Peer-Program Survey so that a reader can check the claim rather than take it.

The survey’s finding is not that peer programs are deficient. It is that the published curricula concentrate, reasonably, on statistical computing in the algorithmic sense: optimization, simulation design, the bootstrap, Bayesian computation. What recurs in the gaps is the operational layer that sits between a method and a result a reader can trust. The command line, SQL and database extraction, principled missing-data practice, dependency-aware pipelines, continuous integration, and de-identification each appear in a minority of published curricula, and each is something a biostatistician will be asked to do within months of starting work. Where a chapter here exists because of such a gap, it says so at its opening and cites the evidence.

Two topics deserve particular mention because their absence has a cost that alumni report directly. The first is CDISC (25  Clinical Data Standards: CDISC, SDTM, and ADaM): graduates entering pharmaceutical and contract-research positions encounter SDTM and ADaM on their first week and find that no course prepared them. The second is SAS (29  SAS for R Programmers), which remains the screening filter for a large share of United States industry positions regardless of what one thinks of the language. Neither chapter argues that these standards are good. Both argue that a graduate who cannot read them is excluded from work they are otherwise qualified for.

Craft is not separable from correctness

It would be possible to read the preceding sections as an argument about professional convenience: learn the tools, get the job, satisfy the funder. That reading understates the case.

The infrastructure in this book exists because analysis is error-prone in ways that infrastructure catches and vigilance does not. A join that silently multiplies rows, a package update that changes a function’s behavior between the analysis and the revision, a figure built from an intermediate that was never rebuilt after the cleaning script changed: none of these announces itself, and none is caught by being careful. They are caught by an assertion, a lockfile, and a dependency graph. The claim that reproducible infrastructure is a courtesy to future readers is true and incomplete. Its more immediate payoff is that it is what lets a statistician carry the ten or fifteen simultaneous projects that the work actually consists of (Slade et al., 2023) without any of them quietly rotting.

At the same time, this book is careful not to claim more for its tools than they deliver. A perfectly reproducible, fully compliant, numerically correct analysis can still mislead every reader it reaches, and no lockfile or container will notice. 34  Ethics Beyond Compliance is the chapter that says so, and it is placed early rather than late because the rest of the book is easier to keep in proportion once the reader knows what the infrastructure cannot do.

What is different here

  • Reproducibility is the organizing principle. Each chapter is framed around producing analyses that a collaborator can rerun and extend. This is the animating concern of modern biostatistics.
  • The judgment is taught explicitly. Every chapter contains a section on what the tool cannot decide for you. These are the decisions that distinguish defensible practice from competent tool use, and they are the part of the work that neither a package nor a model performs.
  • AI-assisted workflows are taught explicitly. Students in 2026 will use large language models in their daily work. Rather than pretend otherwise, the book dedicates a chapter to using LLMs responsibly in R coding, and ends every chapter with adversarial prompts designed to expose model limits.
  • Real case studies. Two full analyses (Palmer Penguins and ADNI MCI prediction) are carried end-to-end, from project scaffolding to final report.
  • Federal context. Biostatisticians working on federally funded research must meet NIH/NSF reproducibility requirements. A dedicated chapter covers what those requirements are and how to satisfy them.
  • The collaborative reality. The biostatistician’s work is done on teams, across several projects at once, with colleagues who cannot read the code. 4  Team Science for Biostatisticians treats this as a skill to be learned rather than a temperament to be possessed.

Prerequisites

Readers should have:

  • An undergraduate statistics background.
  • Basic R familiarity (subsetting, the pipe, writing functions). Readers who lack this should first work through the early chapters of the companion book or R for Data Science (Wickham et al., 2023).
  • Access to R 4.4+, RStudio (optional), and Git.

No prior Docker, renv, or Quarto experience is assumed.

Conventions

See the Conventions page for the visual cues used throughout the book.

Acknowledgments

This book draws heavily on Jenny Bryan’s STAT 545 (Bryan & Stephens, 2019) and Happy Git with R (Bryan, 2019), Ben Marwick and colleagues’ rrtools (Marwick, 2018; Marwick et al., 2018), Karl Broman’s Tools for Reproducible Research (Broman, 2019), Frank Harrell’s R Workflow (Harrell, 2025), and Hadley Wickham’s book family (Wickham et al., 2023; Wickham & Bryan, 2023).

I thank the graduate students whose questions shaped the material substantially.

Ronald “Ryy” G. Thomas
La Jolla, California
Summer 2026