3.6 KiB

Raw Permalink Blame History

2026-03-26

Evening Briefing (8:15 PM ET)

Generated and posted to dashboard. Cron job a954399d-6f5c-4811-9b0f-dc2a4b83833e.

Markets:

S&P 500: 6,477.10 (−1.74% / −114.80) — $900B-$1T equity value erased
Nasdaq: −2.3% (tech-led selloff)
Dow: −1.4%
NVDA: −10%+, META: −7.92% (jury ordered $375M for harm to minors, Reality Labs layoffs)

AI/Tech:

Claude Code v2 released — PowerShell tool (Windows preview), hooks, env vars
Alex Finn: "AI killed dumb software, not all software" — niche problem-solving remains valuable

Trump/Iran:

Reports: Trump weighing ground operation on Iran's Kharg Island oil hub
Trump claims Iran allowed 8-10 tankers through Hormuz as goodwill gesture

Posted to dashboard:

Briefing: id 23a12629
News items: market selloff, Claude Code v2, Trump/Iran, Alex Finn insight, Clarivate AI50

Discord DM sent to johanjongsma with dashboard link.

Rozemarijn (Roos) Study Session — Lecture 7 + Lecture 8

Session: All-day study session, ~10:20 AM – 5:38 PM ET User: Rozemarijn (Discord ID: 1486461895136252115, username: rozemarijn) Exam: Real Estate Research, Thu April 2, 2026, 2 hours, Univ. of Groningen

Lecture 7 — COMPLETED ✅

All slides covered. Key topics:

Functional form (linear, log-linear, log-log) and interpretation of coefficients
Polynomials (concave/convex functions, x²)
Endogeneity — omitted variable bias, dirty x
Durbin-Wu-Hausman test (test for endogeneity)
IV / 2SLS: instrument validity (relevance + exogeneity), two-stage procedure
Weak instrument test: F-statistic > 10 rule of thumb
Interaction variables (city-specific slopes)
Transformations summary table (linear/log-linear/log-log)

Roos' weak areas noted:

Needed multiple explanations of predicted value x̂ (eventually understood with numeric example + analogy)
λ notation confused her (explained it's just Stage 1 β)
Prefers: consistent notation, always show formulas, explain in steps
Can switch to Dutch when struggling

Lecture 8 — IN PROGRESS (slide 17 of 24)

Topics covered so far:

TRUE model (y = x₁β₁ + x₂β₂ + ε)
Restricted vs Unrestricted model (pooled vs separate group lines)
Number of parameter restrictions: DF1 = (g-1)×k, DF2 = n - g×k
Chow test — full calculation with wage differentials example (F=1.30, reject H₀)
What to report after Chow test (RSS + k always, pooled vs separate based on result)
CLRM assumptions A1–A5 (Roos knows A4 well; A1/A2/A3/A5 need drilling)
Consistency vs Efficiency (A4 violation = biased = worst; A2/A3 = inefficient SE)
Slide 17: 4 tests overview (F-test, Chow test, White's test, RESET/functional form test)

Still to cover in lecture 8:

Slide 18: F-test statistic formula (has diagram)
Slides 19-20: White's test for heteroskedasticity (detail)
Slide 21: Cross-sectional data analysis workflow
Slides 22-24: Formative test exercises (Exercise 1 + Exercise 2)

Roos' Notation Preferences (STRICT — always follow)

From her green table image:

E(μₜ) = 0
var(μₜ) = σ² < ∞
cov(μᵢ, μⱼ) = 0
cov(μₜ, xₜ) = 0
μₜ ~ N(0, σ²) Always use μ (not ε or u) for error term when tutoring Roos. Always consistent notation — no variation between explanations.

Mirror To-Do List Created

Full study plan written out per day (March 26 – April 2). Roos has Paradiso performance Fri evening + concert Sat evening (social commitments noted in study plan).

CLRM Drilling Needed

Roos explicitly flagged she needs to know A1–A5 by heart.

✅ A4 — understood well
⏳ A1, A2, A3, A5 — come back to drill these before exam

3.6 KiB Raw Permalink Blame History Unescape Escape