Re:Investigating Jeanne Calment’s Longevity Record
by nonbel on Monday December 17, 2018 @04:32PM
(#18702401)
These types of studies (correlate [common behaviour] with [common health issue(s)] will "discover" whatever the prevailing opinion is of the day.
The structure of the statistical models and what features get included is totally arbitrary, so they just play with it until the "right" result appears. In some cases you can even code male = 1, female = 0 and see (for example) alcohol is positively correlated with a health issue, but then reverse it to be male = 0, female = 1 and see alcohol is negatively correlated (see a simple example in [1]) Totally arbitrary.
The structure of the statistical models and what features get included is totally arbitrary, so they just play with it until the "right" result appears. In some cases you can even code male = 1, female = 0 and see (for example) alcohol is positively correlated with a health issue, but then reverse it to be male = 0, female = 1 and see alcohol is negatively correlated (see a simple example in [1]) Totally arbitrary.
[1] https://news.ycombinator.com/item?id=16714522#16719754