The three-day track
For a cohort with three days · four sessions of ninety minutes a day, the workshop on day 2
The three-day track puts the workshop in the middle of a week and gives it a day on either side. Day 1 builds the tools, so that nobody spends the workshop fighting an installation or meeting R for the first time. Day 2 is the one-day track, unchanged. Day 3 goes back through the methods that the workshop used at speed, in the order in which they were needed, and ends on where research materials live and how work is versioned. The three days share one case throughout: a Multi100 reanalysis of a claim made by the original author (OA) of a paper about how the European Union is framed across its member states. Someone who finishes the week can set up a reproducible project of their own, state the estimand behind a claim and draw the graph that goes with it, fit and read the models that the case needs, and lay out a multiverse of alternatives around a single declared analysis without over-reading what the resulting curve shows.
Each day runs as four sessions of ninety minutes, which is the rhythm of the workshop day itself. Days 1 and 3 are built from the companion curriculum, the nine modules that carry the teaching material of the course, and every module stands on its own, so the internal order of those two days is yours to change. Day 2 has a fixed running order, for the reason given below.
Day 1 – The tools and the reproducible document
Day 1 is about R and about the kind of document that carries an analysis from data to finished report. It ends with the browser lab running and the exercise overview read, so that day 2 opens on its own material with the installation already behind you.
| Work through | What you do | Timing |
|---|---|---|
| Setup and Getting started with R, sections 1 to 5 | Install R and Positron, download the workspace zip and run R/check_setup.R until it prints OK. Anyone whose machine will not cooperate works in the browser lab all week and misses nothing. Then the console against the script, objects and data frames, installing packages, and reading a CSV. |
Session 1, 90 minutes |
| Getting started with R, sections 6 and 7 | Read the pipe left to right, then see one small task on the workshop’s own panel written three ways, in base R, in the tidyverse and in easystats, so that none of the three ever looks foreign. | Session 2, 90 minutes |
| Getting started with R, sections 8 to 10 | The grammar of graphics, built up until you can read the workshop’s own class chart; then naming the quantity you want before you estimate it, which is the habit that day 2 depends on twice. | Session 3, 90 minutes |
| Quarto and reproducible documents | What a .qmd file is and what rendering actually does; the header and the execute options that matter; code chunks and their options; figures, tables and cross-references; citations from a bibliography; and how a document is published. The examples are the files of this site itself. |
Session 4, first 60 minutes |
| Getting started with R, section 11, then Browser lab and Exercise overview | Read the comparison of the two routes into the workshop and settle on yours. Start the R engine in the browser once and run the Task 1 setup check, confirming that the panel has 270 rows. Then read the claim you will work on and what the five Multi100 analysts found. | Session 4, last 30 minutes |
By the end of the day you have a working route into R, one small task written in three dialects, a rendered Quarto document of your own, and the browser lab started once on the panel of 270 country-years.
Day 2 – The workshop
Day 2 is the one-day track, run exactly as it runs on its own: four parts of ninety minutes, breaks of thirty, sixty and thirty minutes, five numbered tasks and two submissions to the class chart. Nothing about it changes because a day sits on either side of it, so the timings, the deck and the run of tasks are the ones on that page, and the table below summarises it. The one thing day 1 buys you is room. A cohort that has already done the R module can drop the optional R tutorial inside Part 2 and give the time to the causal graph instead.
| Work through | What you do | Timing |
|---|---|---|
| The one-day track, Part 1, with Slides and Browser lab Task 1 | Why the field is argued about at all, the twenty-nine teams of the red-card study, the three Rs, Multi100 from the inside, and the toolchain. The setup check runs into the first break. | 90 minutes |
| The one-day track, Part 2, with Browser lab Task 2 and Draw the arrows you believe | The paper in full, estimands, and the causal-graph crash course. You write the estimand in words and finish the graph for the case. | 90 minutes |
| The one-day track, Part 3, with Specification menu, Browser lab Tasks 3 and 4, and Multiverse | Reproduce t = −3.804 exactly, then open the menu of eight axes and explore it. Everyone submits one exploratory result to the class chart. | 90 minutes |
| The one-day track, Part 4, with Reading a specification curve and Browser lab Task 5 | Read the room’s dots against the full curve and the five analysts, find your percentile, work through what a curve can and cannot establish, then declare one analysis, run it and report it. | 90 minutes |
By the end of the day you have reproduced t = −3.804, submitted one exploratory dot and one declared dot to the class chart, and written the report that carries them.
Day 3 – Methods in depth
Day 3 slows down everything that day 2 had to take at speed. The order matters: the graph decides what a model should contain, the models come next, the multiverse is what happens when the choices behind those models are enumerated rather than picked, and simulation is how you find out what a design recovers when the answer is known in advance. The day closes on the two skills that the workshop day names on a slide but never gives you time to practise – where research materials live, and how work is versioned.
| Work through | What you do | Timing |
|---|---|---|
| Draw the arrows you believe | Confounders, mediators and colliders, each one fitted as well as described; graphs built from R formulas with dagify(); the EU-frames graph, whose adjustment set turns out to be what two-way fixed effects already absorb; what fixed effects cannot fix; then a graph for a question of your own. |
Session 1, 90 minutes |
| The statistical methods behind the workshop | Every model that day 2 fitted, output by output: a first linear model read line by line, pooled OLS with cluster-robust errors, two-way fixed effects, random effects and the difference between within and between, and the section where the error structure decides how much evidence a dataset carries. | Session 2, 90 minutes |
| The multiverse: origins, uses, and critics and Reading a specification curve | Where multiverse thinking came from, the three purposes it can serve and why they pull apart, and the 2025–26 critical debate read against this workshop’s own grid. Then the reading: why hundreds of estimates on one dataset are not hundreds of tests, and what a curve-level test would have to do to earn a verdict. | Session 3, 90 minutes |
| Simulation as a method | Data with a recipe you wrote yourself: plant a parameter and see whether the model recovers it, watch the sampling distribution that repeated draws produce, build a multiverse over data with no effect in it at all, and test the precision of a design before running it. | Session 4, 60 minutes |
| Reproducibility repositories and Git and GitHub for reproducible research | Where the materials behind the case live, what a licence lets travel and what it does not, and what registration and forking each mean. Then version control from first principles, and what belongs in a repository against what must never go into one. | Session 4, last 30 minutes |
By the end of the day you have a causal graph drawn for a question of your own, every model that day 2 fitted read output by output, a recovery check on a parameter whose true value you set yourself, and an answer to the question of what may be committed to a repository and what may not.
Adapting this track
Day 1 is the part that compresses. A cohort that already writes R and renders documents can keep only the setup check from it and run the course over two days, and a cohort that writes R but has never met Quarto can keep the fourth session alone. Day 3 compresses by moving reading out of the room rather than by cutting topics: the later sections of the statistical-methods module, from the weights section onward, work well as reading between the sessions.
Day 2 does not compress. The one-day track is already the compressed form of this material, so its own never-cut list holds here in full, and the reasoning behind each item on that list is set out on the one-day track.
A group whose data are individual respondents can put One claim, two levels in place of the simulation session, since it fits the same claim at the individual level and answers the composition question that the aggregate panel can only pose. A group that will publish code together should give the git module a session of its own rather than the half-hour it shares with the repositories module, which is then read as its preparation.
If three days is the wrong shape:
- The one-day track is day 2 on its own, which is how the course is usually delivered.
- The semester track spreads the same material across a term, with the workshop day sitting inside it as the practical core.
- The self-study track drops the clock and the cohort, for anyone working alone.