Digital Mind 🧠

🪵 The Threshold

Epidemiology needs large case counts, so rare-cluster proofs must broaden their outcome measures

Properties1
tagsja-picks, define, seed, craft_and_method, trade, singleton, xref, category/craft-and-method

2 min read

Epidemiology needs large case counts, so rare-cluster proofs must broaden their outcome measures

Evidence

Statistical studies of the sort Zelen specialized in had proved, for example, an irrefutable link between cigarette smoking and lung cancer, but those studies had been based on an analysis of tens of thousands of cases of lung cancer. Statistical studies rely upon large numbers, and Zelen had only twelve cases of leukemia to work with in Woburn. But Zelen thought he saw a way to get larger numbers. He reasoned that if the well water had, in fact, caused leukemia, it might also have caused a variety of other childhood health problems. If an unusual pattern of birth defects and reproductive disorders emerged among families that had gotten their water from wells G and H, that would tend to support the theory that the cluster of leukemia cases was not simply a coincidence, a statistical fluke. — , Harr Jonathan - A Civil Action, 634

Categorization

  • DEFINE: Explains why statistical inference fails on small clusters — the ‘large numbers’ principle of epidemiology, and the workaround of widening the health-outcome net.
  • SEED: Angle: ‘How do you prove a disease cluster isn’t random when you only have twelve cases?’ — the design logic behind surrogate-outcome studies.
  • XREF: connects to: Cancer clusters can be dismissed as random coin-flip groupings. — , Harr Jonathan - A Civil Action , 879; Cluster detection depends on lay residents noticing small run counts. — , Harr Jonathan - A Civil Action , 54; Adding irrelevant cases inflates significance and can flip causal signs — Gary Goertz, Social Science Concepts: A User’s Guide, 1561

Broader Connections

✉️

Backlinks