> Consider a depoliticized example: you do 30 autopsies on people who had recently been to a particular location, let's say a certain grocery store in town. Enough people go to this grocery store that it's not uncommon for way more than 30 recent visitors to be dead unexpectedly by random chance. Further, let's say in the general population, a tiny fraction of people die of a certain food-borne illness. You have a pretty good guess what that fraction is.

It's not just "pick 30 recently dead who went to the store", though.

You've picked 30 people who died after eating at the grocery store's salad bar, and used that as evidence that the salad bar is dangerous, without checking how many skipped the salad bar and died anyways.