Define a reproducible question

Define a reproducible question — guided practice

Specify the population, metric and exclusions before analysis.

Use free local JupyterLab with a Python 3 kernel, or an equivalent notebook environment that can read uploaded local files. Put analysis.ipynb and attendance.csv in the same new practice folder; no extra Python packages are used by the analysis. The CSV rows are S1 north 8 held; S2 south 6 held; S3 north blank held; S4 south 0 canceled; S5 north 7 held; S6 south 5 held; S7 north 4 held; S8 south 9 held. The supplied analyze.py runs the same calculation with Python from this folder. Use only these fictional records.

1. 1. Define the unit

   Write that one row is one session and attended is recorded visits, not unique people.

   You should see: Repeated visitors may be counted more than once across sessions.

   A metric’s unit constrains its interpretation.

2. 2. Define inclusion

   Write include status held with a present nonnegative integer count; exclude canceled and blank attendance with reasons.

   You should see: S3 and S4 will be excluded for different reasons.

   Predefined rules prevent convenient after-the-fact choices.

3. 3. Predict the count

   Count the included session IDs by hand.

   You should see: S1,S2,S5,S6,S7,S8: six rows.

   A small fixture allows an independent prediction.

4. 4. Write a limit

   Add that missing attendance is not estimated and the output cannot count unique people.

   You should see: The question is narrower than total people served.

   A precise limit prevents overclaiming.