Skip to content

Week 1 — Foundations active

June 22 – June 26, 2026 · Chapters 1.1 – 1.5 · Lab 1 · Mentor Session 1 · Problem Set 1 (five problems)

Rebuild probability and frequentist inference from the ground up, using simulation as the microscope, so OLS in Week 2 rests on solid foundations.

By the end of Friday you should be able to write — in code, by hand, and on a napkin — what an estimator's sampling distribution is, why the central limit theorem is the reason any of this works, and why a 95% confidence interval is a statement about the procedure, not the parameter.

Daily schedule

All times Eastern. Morning = independent reading + notebook follow-along. Afternoon = live cohort time on Zoom. 4–5 pm = office hours and stand-up.

Block What Where
9:00 – 12:00 Read Ch 1.1 — Joint, Conditional, and the Two Laws That Run Everything. Work through nb1.1 (LIE/LTV by simulation). Textbook + your fork of the repo
1:00 – 4:00 Kick-off Zoom (1:00 – 2:00) · Pair-debug nb1.1 (2:00 – 4:00) Zoom
4:00 – 5:00 Sub-team stand-up + office hours Zoom breakout
Block What Where
9:00 – 12:00 Read Ch 1.2 — Expectations, Variance, Covariance as Geometry. Work nb1.2 (covariance matrices, correlation heatmaps). Textbook + repo
1:00 – 4:00 Start PS 1.1 + PS 1.2 drills. Async, with Slack channel staffed
4:00 – 5:00 Office hours — Prof. Gao Zoom
Block What Where
9:00 – 12:00 Read Ch 1.3 — Estimators and Their Sampling Distributions. Work nb1.3 (sampling-distribution explorer). Textbook + repo
1:00 – 2:00 Lunch / async —
2:00 – 3:30 Mentor Session 1 — "What is a finding?" (Prof. Gao, all 14) Zoom
3:30 – 5:00 PS 1.3 work block + office hours Zoom
Block What Where
9:00 – 12:00 Read Ch 1.4 — The LLN and the CLT, Shown Not Asserted. Work nb1.4 (LLN/CLT convergence). Textbook + repo
1:00 – 4:00 Lab 1 — Coin-Flip Universe: build a NumPy simulation that shows the LLN by eye and the CLT by animation. Zoom + repo
4:00 – 5:00 Office hours + sub-team check-in Zoom
Block What Where
9:00 – 12:00 Read Ch 1.5 — Hypothesis Testing Done Right. Work nb1.5 (power curves, t-test from scratch). Textbook + repo
1:00 – 3:00 Wrap PS 1.1 – 1.5. Commit, push, open the week-1 PR for review. Repo
3:00 – 4:00 Friday demo — one sub-team presents their cleanest CLT figure. Zoom
4:00 – 5:00 Week-1 retro + Week-2 preview Zoom
5:00 pm Deliverable cutoff —

What to read

The textbook is the primary text. Open the chapter, then the notebook of the same number side-by-side.

Notebooks (notebooks/week-01/ in the repo): nb1.1 · nb1.2 · nb1.3 · nb1.4 · nb1.5. All run headless with numpy.random.default_rng(42); no network access required.

Lab: Lab 1 — Build a Coin-Flip Universe.

Mentor reading for Wednesday: Mentor 1 — "What Is a Finding?".

What to submit (due Fri Jun 26, 5:00 pm ET)

  1. Problem Set 1 — five problems committed to your fork:
  2. Lab 1 notebook (notebooks/week-01/lab1.ipynb) — runs end-to-end on numpy.random.default_rng(42).
  3. Journal entry — one paragraph in journal/week-01.md: what clicked, what is still murky, the one question you would ask Prof. Gao if you could only ask one.
  4. Pull request — open a PR titled week-01 — <your name> from your fork against main. A sub-leader will code-review before Monday.

Reproducibility check

Before you push, run pytest -q notebooks/week-01/ from the repo root. If anything errors out, fix it locally — do not push red.

Data we touch this week

None. Week 1 is all simulated data — that is the point. We need the universe to be one we control so the LLN and CLT are visible, not asserted.

A preview of where real data starts next week:

Stuck? Office-hours playbook

If you are stuck for more than ~20 minutes:

  1. Post in #help on Slack with a minimal reproducible example.
  2. Tag your sub-leader (not Prof. Gao first — sub-leader first).
  3. If still stuck, drop into the 4:00 – 5:00 pm office hour.

Office hours are on Mon, Tue, Wed, Thu at 4:00 pm ET. Friday's 4:00 pm slot is the demo + retro.

Next week preview

Week 2 — The OLS Engine →. Matrix-form OLS, Gauss–Markov, FWL, and the three ways the textbook story breaks (heteroskedasticity, clustering, misspecification). Real CRSP data starts on Wednesday.