Rung 2 — What's Reported

Five dives, built and ready to run. Data in data/, rebuildable with python3 fetch.py.

The rung the whole Academy turns on. The student is no longer looking at a thing — they are looking at a report of a thing, and the gap between the two is the entire lesson.

"This map shows where people complain about rats. Not necessarily where rats are."

Rung 1 taught that a shape is a choice. Rung 2 teaches that the numbers were a choice too, made by whoever decided what to count, when, and who to ask.

Grammar unchanged; only question one escalates:

Q1 the trap. What was actually measured, and when was it recorded? Q2 the finding. What does it say? Q3 the turn. Would you be in this data, or missed by it?

---

2.1 · the proxy, twice — complaints, and the catalogue

These two pair by LESSON rather than by shape, and that is deliberate: rung 2 is about the proxy, and one is a civic proxy, the other a scientific one.

The complaints. Brooklyn 15,099, Manhattan 10,203, Queens 7,213, Bronx 5,963, Staten Island 1,247.

Q2. Which borough has the most rats? The chart cannot tell you. It tells you which borough calls.

The catalogue, and it has a control column built in.

Q1. Did fifty times more earthquakes happen in the 1930s than the 1920s? Look at the second column. A magnitude 5 is felt across the basin — it was counted in 1905 as surely as in 2005, so M5+ is the thing that did not change. It goes 10, 10. M3+ goes 15 to 751.

Caltech's Southern California Seismic Network arrived in the 1930s. The ground did not change. The instruments did.

Then the harder question, and it is the whole rung: the 1990s also spike — 3,243. Is that the instruments too?

No. M5+ spikes with it, 40. Landers in 1992 and Northridge in 1994, with their aftershock sequences. Two spikes in one chart: one is an instrument, one is the Earth, and the control column is how you tell them apart.

---

2.2 · two instruments, one thing — and they are not independent

two_instruments_nyc_rats.csv — NYC, 2024.

The obvious move after 2.1: complaints are only a proxy, so go and check them against a real measurement. The city inspects, and records what it finds.

Q2. Rank the boroughs by complaints. Now rank them by confirmed rat activity. Queens and the Bronx swap. The Bronx is fourth in complaints and third in confirmed rats. The proxy got the order wrong.

And then the finding that makes this the hardest dive in rung 2.

Q1. Where do inspections come from?

The city inspects because somebody complained. So the second measurement was caused by the first one. Across the five boroughs, complaints and inspections correlate at r = 0.907.

You cannot check a proxy with a measurement the proxy triggered.

The last column is where that becomes visible: Queens gets 2.3 inspections per complaint and the Bronx gets 8.0. That ratio is not a fact about rats. It is a fact about how the city answers a phone call.

Q3. If your block has a rat problem and nobody on it calls 311, what happens to your block in every chart on this page?

---

2.3 · a pattern with a mechanism, beside one without

Q1. One of these subjects has a week. Which one? Q2. Describe the SHAPE of each pattern, not its size. Q3. Which one would you bet repeats next year?

The sizes are not that different. The shapes are.

Rat complaints fall smoothly from Monday to Saturday — a gradient, in order, every step. That is what a mechanism looks like: people are at work, at a desk, near a phone.

Earthquakes have no gradient at all. Wednesday happens to be high. The Earth does not know it is Wednesday, so that bump is chance, or it is one swarm of aftershocks landing on a Wednesday and dragging the bar up.

A difference in magnitude is not a finding. A difference in shape is.

Teacher note: the quake bump is larger than a naive coin-flip model predicts, and the reason is worth knowing even if it is not taught here — earthquakes are not independent events, they arrive in sequences, so the ordinary "is this noise" arithmetic understates the wobble. The honest answer to Q3 is "the rats, and I am not sure about the Wednesday."

---

2.4 · the outage — the hole and the spike are one event

hourly_311_outage_2022_09_24.csv — 72 hours, LA, 23–25 September 2022. Daily context: hist_la_311_per_day.csv.

Two stages, and taking their answers first is the whole design.

Stage one — the daily chart. Saturday 2022-09-24 shows 95 requests against a 3,638 average. Sunday shows 3,716, the busiest Sunday of the year.

Q2. What happened that weekend? Take every explanation. They will be about people — a quiet Saturday, a busy Sunday, a holiday, an event. Write them on the board.

Stage two — the same weekend, by hour.

`` Fri 23rd normal all day … 19:79 20:47 21:33 22:0 23:0 Sat 24th 00–18: NOTHING 19:1 20:5 21:25 22:38 23:26 Sun 25th busiest Sunday of 2022, and the volume runs LATE ``

Nineteen consecutive empty hours. The system stopped at 21:55 Friday and came back at 19:49 Saturday.

Q1. Nobody reported a single thing for nineteen hours and then started again at a quarter to eight in the evening. Is that people? Q3. Which of the explanations on the board survives?

The dip is instrumentation. And so is the spike — Sunday is the backlog landing. One chart, two features, one cause, and the cause is not in the chart. It is in the hours, which the daily chart threw away.

It happened twice: 2022-03-26 carries the identical signature, seventeen empty hours, back at 19:00. Full workings in SOURCES.md.

---

The MAKE dive

They have spent five sessions finding out that numbers were collected by somebody.

Design the three questions you would ask to find out how the Academy is doing. One answerable in one word. One needing a sentence. One about something that did not happen.

Then, before you ask anybody: who will not answer, and what will that do to your result?

That is the Handoff's Phase 0 collection design (courses/is1/units/the-handoff/), reached by a class that now knows why the third question exists.

---

Not built, and why

Provenance

SOURCES.md. Publishers this rung adds: NYC OpenData (311 Service Requests; DOHMH Rodent Inspection). Everything else was already banked.