Lecture Platform
Mini-project B — optional, non-gating Fishing trip

Take a random 300-visit subsample. Run the readmission-rate test across all 14 available categorical columns (race, gender, age band, admission type, discharge disposition, admission source, specialty, payer, glucose test, A1C, medication change, diabetes medication, metformin, insulin). Count how many come out “significant” at p < 0.05. Then re-test only those survivors on the other 101,466 visits — the ones that vanish were fish.

Visual

p-values before (300-visit sample) vs after (held-out re-test) for the columns that first looked significant
4 of 14 columns looked "significant" on the 300-visit sample; re-tested on the other 101,466 visits, only 2 survived — the other 2 were fish.

This is ASA Principle 4 (selective reporting distorts the interpretive value of p-values) turned into a lab exercise, and it's the same train/test idea Module 03 opens with. Open mini-project.ipynb to reproduce it yourself.

This marks "multiple-comparisons" as touched in your progress tracker. It does not gate module mastery — Project A's six checkpoints are the only requirement for "project complete."