WE CAN LEARN MORE ABOUT LONGEVITY FROM THE ELDER THAN FROM THE ELDEST (thread)
Yesterday, I read about a 392-years-old shark
Molson, in the thread below, correctly points out:
- can be studied for longevity BUT
- probably that shark is not that old
1/N
Yesterday, I read about a 392-years-old shark
Molson, in the thread below, correctly points out:
- can be studied for longevity BUT
- probably that shark is not that old
1/N
2/ I do not know much on carbon dating, but I know something about statistics, and I see a problem with the approach of studying the oldest shark.
How do we know it is the oldest shark?
We use a carbon dating model
And how do we build the model?
Calibrating it with sharks
BUT
How do we know it is the oldest shark?
We use a carbon dating model
And how do we build the model?
Calibrating it with sharks
BUT
3/ A general rule is that,
Models created through calibration are imprecise in estimating outliers.
Models created through calibration are imprecise in estimating outliers.
4/ Technical note: yes, this is only true in Extremistan, and age (should) reside in Mediocristan, but:
- what about the distribution of isotopes.
- 392 yo makes the distribution large enough to justify the statement that a model calibrated with a few young sharks is imprecise.
- what about the distribution of isotopes.
- 392 yo makes the distribution large enough to justify the statement that a model calibrated with a few young sharks is imprecise.
5/ We see a similar phenomenon with humans. Many of the oldest people (110+ yo) tend to have a complicated anagraphic history with missing birth certificates or similar sources of uncertainty.
You have lower uncertainty on age on a old person than on the oldest.
You have lower uncertainty on age on a old person than on the oldest.
6/ The general principle of Wittgenstein's ruler (thread below) holds that the more the free parameters in a measurement, the less we know what we are measuring.
Now, the more we measure something old, the less we know if the correlations we base the measurement on are correct.
Now, the more we measure something old, the less we know if the correlations we base the measurement on are correct.
7/ For example, all other things equal, the more we go back in time, the less birth certificates are reliable and the less we know whether carbon dating based on extrapolations from calibration is correct.
8/ Therefore, my guess is that:
- Yes, studying the elder might provide some insights towards what drives longevity.
- Studying the eldest is much, much trickier. Samples are smaller, uncertainties higher, and conclusions less robust.
- Yes, studying the elder might provide some insights towards what drives longevity.
- Studying the eldest is much, much trickier. Samples are smaller, uncertainties higher, and conclusions less robust.
9/ In conclusion, to solve the problem of the outliers above, let's study the top 1% oldest, not the top 0.001%
More in general, it is probably easier to increase longevity by clipping the left tail of the age distribution than by increasing the right tail IMHO
(explanation 👇)
More in general, it is probably easier to increase longevity by clipping the left tail of the age distribution than by increasing the right tail IMHO
(explanation 👇)
10/ This means, easier to find and apply ways not to die early than to find and apply ways to increase the maximum potential age.
11/ This is also because we have better data, we are more certain of causal links, and have lower uncertainty due to environmental changes related to how to clip the left tail than how to increase the right tail.
