4 ms·

Simpson’s paradox is when a trend or correlation is observable in each of the sub-populations, but vanishes when the data is aggregated. For example, a drug tha

Simpson’s paradox is when a trend or correlation is observable in each of the sub-populations, but vanishes when the data is aggregated. For example, a drug that has a strong effect on men and women when analyzed separately, but shows little effect at the population level. This example is not Simpson’s paradox, it is simply the misuse of statistics. Statistics, being mechanical transformations of data, only have semantics within a causal model. Simply picking variables randomly and then assuming causality when the statistics behave that way is inverting the process of knowledge formation. EDIT: Thanks for the corrections—the real data that this fictional example is based on does show Simpson’s paradox, as the dependent variable (death rates) appears to show a positive correlation with vaccination status when aggregating the population, but a negative correlation for every age group individually.

lordnacho 5y ago
Isn't that what's happening here? There's a guy who claims vaccination has the opposite effect to what people normally claim, that they actually cause people to die. And it seems to be because people from age 10 to 59 have been aggregated. The sub-populations, for instance ages 10-19, 20-29, etc would not be showing that vaccinated individuals are dying more often. It's only by aggregating them you get the wrong conclusion.
- jkhdigital 5y ago
  You’re right, I did read to the end of the post to see that but for some reason I fixated on the “contrived” example and the fallacy that it entailed.
theli0nheart 5y ago
> This example is not Simpson’s paradox, it is simply the misuse of statistics. If you read to the end of the post, you'll see that the author was using this correlation to prove that the mistake is identical to another claim related to COVID [1]. This COVID-related correlation doesn't seem as spurious as the Ghostbusters one, but that's because it's much harder to spot errors like this when variables aren't so "random". [1]: "Vaccinated English adults under 60 are dying at twice the rate of unvaccinated people the same age"

lordnacho 5y ago

Isn't that what's happening here? There's a guy who claims vaccination has the opposite effect to what people normally claim, that they actually cause people to die. And it seems to be because people from age 10 to 59 have been aggregated. The sub-populations, for instance ages 10-19, 20-29, etc would not be showing that vaccinated individuals are dying more often. It's only by aggregating them you get the wrong conclusion.

jkhdigital 5y ago
You’re right, I did read to the end of the post to see that but for some reason I fixated on the “contrived” example and the fallacy that it entailed.

theli0nheart 5y ago

> This example is not Simpson’s paradox, it is simply the misuse of statistics. If you read to the end of the post, you'll see that the author was using this correlation to prove that the mistake is identical to another claim related to COVID [1]. This COVID-related correlation doesn't seem as spurious as the Ghostbusters one, but that's because it's much harder to spot errors like this when variables aren't so "random". [1]: "Vaccinated English adults under 60 are dying at twice the rate of unvaccinated people the same age"