3 ms·
I think this is basically Simpson's Paradox. If you take an average across the entire population, that is actually the average if you were to randomly select a
by alexpetralia 3y ago
I think this is basically Simpson's Paradox.
If you take an average across the entire population, that is actually the average if you were to randomly select an individual from the population (i.e. the long-run value after repeated sampling).
However, if you were to examine particular subgroups, they may wildly deviate from the average, and so in any particular subgroup, the mean is not at all like the population average.
If you have more information on the particular subgroup you are sampling from, then you should of course use that information to estimate the relevant average. If you don't, or the claim you're making spans many comparable subgroups, then perhaps a broader average applies.
- lucas_membrane 3y agoI saw a report recently that said that a population study (from some country in Europe, IIRC) showed that people who are still getting their COVID vaccinations updated are subsequently having worse COVID experience than others. Unfortunately, there are three possible explanations for this that immediately come to mind: (1) The vaccine increases risk; (2) People have some way of inferring their need for vaccine, which the study could not or did not take into its analysis, and the vaccinated group was a self-selected high risk group; (3) The vaccinated group felt protected enough by the vaccine to eschew isolation and make itself a high risk group. Note that 2 of these work in one direction an 1 in the other, and that all three might simultaneously be true enough of some subset of the studied population to be significant in determining the study's results. Do science, but carry humility, especially about why people do anything, what that means, and what is unknowable within your context.