Teacher's guide

The Tour, for a first session

The 45-minute tour visits ten key figures, from the course's 100 patients to a targeted survival curve. Each stop links straight into its lesson, and each of those figures appears on that lesson's page below under “Key figure”.

Project it in the first session as a map of the whole course, or send it as pre-reading. For a single talk, a journal club or grand rounds, the Tour is the talk.

The placement quiz, for who starts where

“Where should I start?” on the course map asks eight short questions, each with an “I'm not sure” answer, and recommends one of six starting points (the beginning, Design, the payoff, the geometry, estimation, or Trust the answer) plus a few lessons to review.

It runs in the student's browser and sends you nothing, so ask students to write down their result. Its rules were written by hand from the course content and have not been validated as a test: use it to start a conversation and to group students, not to grade them.

A first session that uses both

• 10 minutes: everyone takes the placement quiz and notes the starting point.
• 25 minutes: project the first five Tour stops; stop at each figure and ask its question.
• 15 minutes: pairs that placed differently explain one Tour figure to each other.
• 10 minutes: agree the schedule. Students who placed later may skip earlier sessions, but should work through the review lessons the quiz lists.

What each lesson page gives you

Goals, the key figure with questions to ask while it plays, three discussion questions with a collapsible answer key, one in-class activity with materials and what good work looks like, the misconceptions students often bring, and a slide-free lecture outline that walks through the lesson's own figures.

Answer keys are hidden when you print, unless you tick “Show every answer key” at the top of this page.

Who the course is for

Applied trial statisticians and epidemiologists who already use regression, Kaplan–Meier, Cox and propensity scores, and want to see what AIPW and TMLE are actually doing. It also suits graduate students in biostatistics or epidemiology who have met those tools but not the theory behind influence functions.

The route runs from the causal question, through design and the geometry of estimation, to intervals you can defend and back to survival curves. The regulatory thread (ICH E9(R1) estimands, target trial emulation, FDA guidance on covariate adjustment and on device real-world evidence) runs through it, so it works for industry and academic groups alike.

What students need first

• Linear and logistic regression, and a confidence interval.
• Kaplan–Meier curves and the Cox model, as users.
• Propensity scores and inverse probability weighting, at least by name.
• Expectation and variance, and the derivative of a function of one variable.

No measure theory. Every derivation is built from sums students can see. R is optional: several lessons carry a short base R script with its verified output, and each lesson page below says which.

How to use it in a class

Each lesson is a sequence of steps, one idea per step, with a figure that moves and a prediction to make before the answer appears. Students can enter any lesson directly; each lesson says what it builds on.

Every lesson opens with Learning and display settings. Guided shows one step at a time and can reveal one idea at a time; Explore shows every step at once, which is the better mode for projecting in class. Laboratory lessons have a Share this configuration button: set up the figure the way you want to discuss it and send the link.

A flipped classroom

The lessons are built for self-study, so class time can go to talk. A session that works:

• Before class, students complete the lessons for that session (the times below are estimates of working time).
• 10 minutes: retrieval. Each lesson opens with two questions from the one before; ask them cold.
• 20 minutes: project the lesson in Explore mode and follow its slide-free lecture outline.
• 15 to 20 minutes: the in-class activity, in pairs or small groups.
• 10 minutes: one discussion question, then the lesson's “Try a new case” as an exit ticket. Use the transfer check as an exit ticket, not as the activity.

What progress means, and what it does not

Progress lives in each student's browser, on their own device. There are no accounts and no server, so you cannot see your students' progress; ask them to show you.

Each lesson ends with a transfer check: a new numerical case with a hint and a worked solution. A lesson moves through explored, attempted, practiced with assistance (a correct answer after revealing that case's solution) and transfer check demonstrated (a correct answer on a case attempted independently). A new case can always be drawn. Scrolling and revealing answers never demonstrate a lesson. A successful check schedules a reminder on the course map three days later.

Open licenses

The code is under the MIT license (LICENSE). The lesson text, explanations, analogies and worked examples are under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 (LICENSE-CONTENT.md). You may teach with them, share them and adapt them, with attribution, for non-commercial purposes, under the same license.

Lesson times come from the course map and are estimates of working time (reading plus interaction), not measured completion times.