Hacker News Reader

This isn't a good control if the record accuracy depends on the age of person.

Imagine if you're trying to find medical fraud and you find out that a lot of supercentenarians have cancer. You try to control by comparing them to the average, and supercentenarians have higher rates of cancer than normal. Is that fraud, or does age lead to higher cancer rates?

View on HN

The parent suggested controlling by the date of birth of all people born same year. I.e. if only 20% of all people born in 1900 have DOB registered on the first of month, but 50% of supercentenarians born in 1900 have DOB at the beginning of the month then something is fishy. The records accuracy should be same.

Yes, exactly. The question is if having a DOB on the first of the month grants mystical longevity. This can be answered regardless of the general accuracy of birth records.

That is a decent control, though I think that finding "all people born in 1900" is probably more difficult than you'd imagine. I propose adding "who died 20-30 years ago" because that will make it easier to find comprehensive data. You would need to make sure your survey isn't biased towards people who were notable enough at age 90 to be included, because living past 100 makes people notable in a different way.

Reply on news.ycombinator.com