I originally typed something like 11, based on my guess at the list of people who I personally know are still active and fit the description. Then I decided that was probably unfair, so in the spirit of risk assessors everywhere, I just multiplied it by an arbitrary number.
I am not sure it could get much worse than the NYT Health section always was. (Not sure if that even still exists, but it was always terrible. Always a sure source of an example to grab to walk into class with and ask the students to identify what is wrong.)
In fairness to the epidemiology estimates, most of them are presented in multiplicative terms like 27% reduction, rather than an additive like "reduction in risk of 2 percentage points" or "increases your life expectancy by 0.8 years". So if you keep multiplying those reductions from health-promoting actions you get close to zero (which is still quite possibly absurd), but do not get a clear mathematical error. However, you are right that some of the results are presented in terms of something like life-expectancy increase, and I recall a couple of amusing papers from back sometime that gathered a bunch of those together and added them up to produce the inevitable absurd result.
(Incidentally, that is not the reason the results are almost always multiplicative. Nor is it because there is a good theoretical or empirical reason to believe that is the right functional form. It is because the only statistical methods most people doing epidemiology know how to use inherently assume the relationship must be multiplicative and thus that is the only answer they can present. This is often a big problem for understanding when decision-relevant information is in absolute (additive) terms. So, a 50% reduction in risk for some outcome sounds huge, but if the risk for that outcome was only 2 in 100 to start with, the worldly implications are not actually so huge.)
Really informative. I leave this post feeling less stupid! Thanks!
Bravo! You are among the 37.
I originally typed something like 11, based on my guess at the list of people who I personally know are still active and fit the description. Then I decided that was probably unfair, so in the spirit of risk assessors everywhere, I just multiplied it by an arbitrary number.
The broccoli example is just about perfect. If you took all the "reduce your rate" advice that comes out every day, you would statistically never die.
;-)
In a world of AI Slop and worth being measured by clicks and shares, I expect things to get worse.
I am not sure it could get much worse than the NYT Health section always was. (Not sure if that even still exists, but it was always terrible. Always a sure source of an example to grab to walk into class with and ask the students to identify what is wrong.)
In fairness to the epidemiology estimates, most of them are presented in multiplicative terms like 27% reduction, rather than an additive like "reduction in risk of 2 percentage points" or "increases your life expectancy by 0.8 years". So if you keep multiplying those reductions from health-promoting actions you get close to zero (which is still quite possibly absurd), but do not get a clear mathematical error. However, you are right that some of the results are presented in terms of something like life-expectancy increase, and I recall a couple of amusing papers from back sometime that gathered a bunch of those together and added them up to produce the inevitable absurd result.
(Incidentally, that is not the reason the results are almost always multiplicative. Nor is it because there is a good theoretical or empirical reason to believe that is the right functional form. It is because the only statistical methods most people doing epidemiology know how to use inherently assume the relationship must be multiplicative and thus that is the only answer they can present. This is often a big problem for understanding when decision-relevant information is in absolute (additive) terms. So, a 50% reduction in risk for some outcome sounds huge, but if the risk for that outcome was only 2 in 100 to start with, the worldly implications are not actually so huge.)