# Chapter 1 · Classroom ledger investigation

**Constructed teaching data · developmental answer notes.** The CSV has four rooms, ten recorded intervals per room, and 1,000 occupied seat-hours. No observation in this file establishes a ventilation fault or a health outcome.

The room rates are 20%, 40%, 60% and 80%. Their unweighted mean is $(.2+.4+.6+.8)/4=.50$. The direct seat-hour rate is $(20+80+180+320)/(100+200+300+400)=600/1000=.60$. The two answers are not contradictory: one gives each room equal weight, while the other gives each recorded occupied seat-hour equal weight.

After removing room D, the room mean is $(.2+.4+.6)/3=.40$. The seat-hour rate is $(20+80+180)/(100+200+300)=280/600\approx.4667$. If room A then doubles its recorded occupied seat-hours while retaining its 20% hot fraction, the new seat-hour rate is $(40+80+180)/(200+200+300)=300/700\approx.4286$. Changing a weight can change a weighted answer without changing any room's rate.

Before writing a headline, state the population (for example, occupied seat-hours in the recorded rooms during the defined observation window), unit, temperature threshold, denominator, missing interval policy and decision. The 20% line in the chapter is an illustrative inspection rule chosen by people, not a property discovered by the file. A different threshold requires a new declared rule and perhaps new data about comfort or risk.

The local Python program was executed against the retained CSV and produced:

```text
Room mean: 50.0%
Student-seat-hour rate: 60.0%
Without D — room mean: 40.0%; seat-hour rate: 46.7%
Without D, A doubled — seat-hour rate: 42.9%
```

**Audit prompt.** What was not logged? A sensor beside a hot computer, absent readings during an outage, changing occupancy and a poorly chosen threshold could make the table's recorded rate differ from the quantity the school actually needs. The CSV alone cannot resolve these possibilities.
