logoalt Hacker News

BoxFour01/21/20251 replyview on HN

Why would that be a problem for reporting relative results if the entire population is doing that?

If everyone is under-reporting their alcohol consumption, that seems fine. The absolute numbers will be way off, the relative numbers to their peers won't.


Replies

bluGill01/21/2025

Statistics can do a lot to find data from noise like this, but it is still noise. The biggest issue is nobody knows what variables are important, which are correlated, and so on.

Edit: there is another issue I forget until now: time. Statistically I have several more decades of life left. So even if you get accurate results of my meals yesterday, you need to report when I died, and you probably won't have the meals for the rest of my life. Did some meal I at when I was 10 have a big effect on my life? For that matter if I know you are tracking just one day's meals I will probably eat what I think is better and that doesn't tell you anything about what I eat the rest of the time.

It is easy to track people who have had a heart attack - they are likely to die of another heart attack in a few years so the study times are short. However does having had a heart attack mean either genetic difference such that your results only apply to a subset of the population, or perhaps some other factor of having had a heart attack.