Instructor plan · HEG Genève · International Business Management · Applied Statistics · 2026–27

A flipped, data-driven statistics course

Each session is three hours in two parts: about an hour of lesson, then two hours of exercises worked with the room. The slides and the site mirror each other section by section, so you present and they follow; and the instructor dashboard tells you, each week, exactly where the class is stuck. This plan follows the official syllabus — sixteen weeks, Saylor's Introductory Statistics, 70 / 15 / 15 — and borrows its supervision model from the E1410 course: one running project, earned progress, everything visible before the next session.

Course hub →Dashboard →

The core move: before each session, open the dashboard and look at which sections stalled, which self-check questions scored lowest, and which workshop fields are empty. Aim that week's class at those gaps. A statistics course taught with live data about the class itself is the whole thesis, lived.

1. Assessment — the syllabus, and how the site evidences it

The syllabus prints 70/15/15; this course runs 60/20/20, moving weight onto the two components that are built week by week. What the site adds is evidence: presence and the project are both observable as they happen rather than reconstructed at the end.

60%

Final exam · Week 16

Individual, 90 min, closed book. Authorised: a simple non-programmable calculator and two hand-written sheets (two-sided). Six exercises × 10 points, graded to the half point. Every week's self-check quiz and the exam cards rehearse exactly its archetypes.

20%

Individual project

The pizzeria file: every week's workshop page asks the student to apply that week's method to their pizzeria, and the answers are saved. The project is that file, finished and argued. The dashboard's assignment reader shows it per student, with the gaps spelled out.

20%

Presence · pro rata

A button on the week page, grey until you open the window in the room. Students press it once; you close it. The dashboard grows a column per session and computes the rate — which is the mark. Click any cell to correct it by hand.

Homework · every week

Assigned weekly, not graded, with hints and worked solutions. Scores still reach the dashboard, which is how the weakest problem of the week is found before the session.

Two safeguards, so the project and presence marks are defensible:

First attempt wins. Quiz and classify picks are recorded once and replayed on reload; a student cannot overwrite their own record. The dashboards show the honest first number. built ✓
Per-student randomised numbers in the practice arena and exam cards — copying is pointless. built ✓
Time-on-task sanity. The dashboard logs active minutes; near-zero time with full completion is a flag, and the weekly in-room checkpoint is the cross-check.
The blank test (Week 15) is taken at home and corrected in class — informative, never graded, and the moment students discover what their two sheets are missing.

2. The repeatable session — three hours, two parts

Pre-class: students open the week page. Part 1 is the lesson, presented from the deck; every slide names the section of the site it corresponds to, and every section names its slide, so a student who looks up can always find the room. Part 2 is the exercise set, worked together: the same exercise is on your slide and as a form on their screen, and you release each solution when the room is ready.

Time	Block	Why it needs the room
Part 1 · the lesson · ~60 min
10 min	Retrieval opener — the two questions about last week at the top of the week page, done live; then the dashboard's worst self-check question from last week, argued in pairs and re-voted.	Repairs the misconception before anything new is built on it.
45 min	The lesson, presented — the deck (▶ Lecture on the week page), one "On the pizzeria →" callout per concept, the labs run live on the big screen. Students follow the numbered sections on their own device; each slide names its section.	The argument, the tempo, and the questions a page cannot answer.
5 min	Break and set-up: calculators out, the exercise section open on every screen.	Three hours needs one.
Part 2 · exercises, worked together · ~100–110 min
95–105 min	Eight exercises. Each goes up on a slide and is on their screen as a form. They attempt → Check (numeric answers mark instantly) → Hint. When the room is ready you press Show solution: it appears on the projector and unlocks the solution on every student device. The last exercise of each set is written, not numeric — they compare with a model answer and with each other.	Worked practice with someone to ask, at the pace of the slowest third — which is the thing a screen alone never gets right.
10 min	Checkpoint and close — the self-check quiz on the site (scores land on your dashboard live), the two worst questions reviewed, and the week's workshop entry started.	Accountability for the flip, and the presence evidence.

What you see that they do not. On a device where the instructor gate is unlocked — the same password as the dashboards, remembered per device — every solution is in the speaker notes from the start, the exercise screen shows a 👁 Instructor view strip with a Release all button, and each exercise carries its own Reveal to class. On a student device the solution text is never rendered into the deck at all, and the site's Solution button stays locked until you release it. Releases are permanent, so the whole set is open for revision afterwards. The student sign-in prompt is also suppressed on your device, so it cannot land on top of a slide mid-lecture.

Hours per week, and the semester total

Week	Part 1 · lesson	Part 2 · exercises	Checkpoint & close	Self-study (site + homework)
1–12 · each content week	1.0 h	1.75 h	0.25 h	≈ 3 h
13–14 · exercises	0.5 h	2.25 h	0.25 h	≈ 3 h
15 · blank test	—	2.5 h correction	0.5 h	≈ 2 h (the paper itself)
16 · exam	90 min written exam			—
Semester	≈ 13 h	≈ 28.5 h	≈ 3.5 h	≈ 44 h

Contact time ≈ 45 h over fifteen 3-hour sessions plus the exam; self-study ≈ 44 h. The project's own writing time (≈ 10 h) is additional. Roughly a third lesson, two thirds practice — which is the ratio the exam rewards.

3. The sixteen weeks

The syllabus order, with the Saylor chapter, what students do on the site before class, what happens in the room, and the build status of each week's page. Weeks 1–2 are complete; 3–16 are scaffolds — real pages with the plan, objectives and slide outline, openable from here for review, locked on the student hub until written.

Wk	Topic (Saylor)	Pre-class (site)	In the room	Status
1	Introduction — before the formulas ch. 1	Week 1 · HW1	Course launch; lock in a price; "model not truth"; the four traps; checkpoint live.	BUILT · 9 sections · 27 slides · 8 exercises (95 min) · 10-q quiz · 4 workbook fields
2	Descriptive statistics ch. 2	Week 2 · HW2 · arena	Histogram live; the n − 1 trap; box plot by hand; defend a price; checkpoint.	BUILT · 9 sections · 25 slides · 8 exercises (102 min) · 10-q quiz · 5 fields
3	Discrete random variables ch. 4 (+ ch. 3 rules)	Week 3	Two-way count probabilities; a distribution built by dragging; the binomial; P(sell-out).	SCAFFOLD · 7 sections · 13-slide outline · exercise outline
4	Continuous random variables ch. 5	Week 4 · distributions	Area = probability; the z table; standardise delivery times; the inverse.	SCAFFOLD · 7 sections · 12-slide outline · exercise outline
5	Sampling distributions ch. 6	Week 5 · simulator	The keystone: hands-on sampling (candy/dice) and the simulator; σ/√n; the CLT; p̂.	SCAFFOLD · 6 sections · 11-slide outline · exercise outline
6	Estimation ch. 7	Week 6 · catch the mean	Build a CI for the class data; the CI-misinterpretation clinic; z or t; sample size.	SCAFFOLD · 7 sections · 11-slide outline · exercise outline
7	Testing hypotheses ch. 8	Week 7	The p-value misconception clinic; a permutation test by shuffling labels; Type I vs II.	SCAFFOLD · 7 sections · 12-slide outline · exercise outline
8	Two-sample problems ch. 9	Week 8	Design and run a class A/B test; paired vs independent — the decision.	SCAFFOLD · 7 sections · 11-slide outline · exercise outline
9	Correlation and regression ch. 10	Week 9	Fit a line to class data; the slope test; the lurking variable returns.	SCAFFOLD · 7 sections · 12-slide outline · exercise outline
10	Chi-square tests and F-tests ch. 11	Week 10	Contingency-table activity; ANOVA; the "which test?" decision-tree drill.	SCAFFOLD · 7 sections · 12-slide outline · exercise outline
11	PCA and other non-parametric tests DataTab	Week 11	Computer session: Mann–Whitney and a PCA in DataTab on the customer survey.	SCAFFOLD · 7 sections · 12-slide outline · exercise outline
12	Statistics in Python Colab	Week 12	Computer session in Google Colab: the semester in twelve lines of Python; read an output without panic.	SCAFFOLD · 6 sections · 10-slide outline · exercise outline
13	Exercises and repetition I weeks 1–6	Week 13 · cards	Timed exam-style exercises 1–3, worked live; chosen from the dashboard's weakest questions.	SCAFFOLD
14	Exercises and repetition II weeks 7–12	Week 14	Exercises 4–6; the "which test?" tree; interpretation drill.	SCAFFOLD
15	The blank test	Week 15	Correction of the take-home paper; what the two sheets must contain.	SCAFFOLD · paper to write
16	Final exam	Week 16	90 min, closed book, six exercises. Project hand-in on Cyberlearn.	SCAFFOLD · info page

4. Supervising students — the E1410 model, applied here

Four instructor views, all reading the same data the week pages write. Nothing is typed twice and nothing is reconstructed at the end of term.

Dashboard — what do I do before the next session? Ranked actions, the cohort's shape, one roster table with a per-student drawer: sections done, time on task, self-check scores, the pizzeria file so far. Flags are relative to the cohort's own median, never a fixed threshold.
Wide table — every student's state on every week and on the project, in one scrollable table with sticky names. The per-question quiz stats live here: the two worst-answered questions of last week are the retrieval opener of the next.
Insights — where is the course failing them? Section-level stall points, week-to-week drop-off, workbook fill rates — including which classify items misled people, with the scenario text.
Messages — the student panel is on every page (bottom-right); the instructor's end shows threads with one student, a group or the whole class. Announcements post from the dashboard and appear as a banner on every page.

Identity: students sign in once with their name and a six-digit code (/shared/login.js), so progress follows them between laptop and phone and reaches the namespace statistics. Progress is always written locally first and back-filled on sign-in, so nothing done before signing in is lost. Lost codes are on the roster in the dashboard. The site is public and listed; there is deliberately no class password.

5. Build backlog

What exists, and what each remaining week needs, in build order. The pattern is fixed by Weeks 1–2: eight earned sections, one hook estimate, classify exercises, a lab per concept, a 10-question checkpoint, a workbook, a 15–17-slide deck with speaker notes and an "On the pizzeria →" thread.

DONE — HEG theme (/shared/themes/heg.css), the week-page stylesheet (stats.css), the hub, the course-progress bar, Weeks 1–2 with homework, redirects from the old page names, the seven tools restyled.
DONE — the in-class exercise engine (exercises.js): one array per week renders the form, the slide and the speaker-note solution, with instructor-released solutions. Weeks 1–2 have full sets (8 exercises each).
DONE — scaffolds for Weeks 3–16, generated from _build/scaffold.py: each now carries a Part 2 exercise outline (title, minutes, what it drills) as well as the section and slide outlines. Rerun after editing the plan; delete a week's entry once its page is hand-written.
NEXT — Week 3 (discrete RVs) and Week 4 (normal): reuse the exam-card generators c3–c6 for the labs; the distributions explainer already carries the pictures. Each week now needs two things written — the six lesson sections and the eight exercises, whose outlines are already in the scaffold.
NEXT — a per-week answer-key check in _build/: recompute every ans: from the dataset in Python, the way Weeks 1–2 were verified, so no wrong number ever reaches the projector.
NEXT — Week 5: embed the simulator in the page rather than linking out; Week 6: a CI-capture animation from catch the mean.
NEXT — Weeks 7–10: the OMBA401 weeks 5, 6, 7 and 9 carry working labs for tests, ANOVA, regression and chi-square (with exact p-values via logGamma / incomplete beta) — port the maths, rewrite the case for the pizzeria.
NEXT — homework 3–12 on the shared engine; exam cards c7–c27 from "soon" to playable, week by week as each is taught.
NEXT — a compile.html that assembles the pizzeria file in project order (the E1410 compile.html is the template; courses.json → project.sections already lists the fields), and the Week 15 blank-test paper.
LATER — Week 11 DataTab walkthroughs with screenshots; Week 12 Colab notebooks (one per section, shareable links); a French chrome if the cohort needs it.

Living document · adjust freely. Course hub: HEG/statistics/index.html · Dashboard: admin2 · Generator: HEG/statistics/_build/scaffold.py