9 Selection Effects
1) Modern domestic cats are largely neutered, to prevent them having litters of their own. Street cats, of course, are less-likely to get neutered, because they're out in the streets. This means that if you adopt a cat from an adoption shelter (or similar), they're more likely to be the descendent of a street cat, who (on average) got there because they didn't like humans, or homes, or comfort. Over time, the population of cats is being selected for disdaining humans.
2) When I was in university, a vast share of my Indian-American friends were vegetarian. Without really thinking about it, I assumed that Indians in general were overwhelmingly vegetarian. When I eventually went to India, I found out this wasn't exactly true. What I now think happened is that certain castes of Indians are more likely to be vegetarian, Indians in America are disproportionately likely to be from those castes, and Indian-Americans at my university were even-more-disproportionately likely to be from those castes. (I haven't fact-checked these numbers carefully but the internet says 40% vegetarian rate for Indians in India vs 60% for Indian-Americans).
3) in order to become President (or reach other high office), you usually must have previously Proved The Haters wrong multiple times beforehand. Joe Biden had been running for president and losing for 30 years before he succeeded; Donald Trump's candidacy was treated as a literal joke by almost-everyone, until it wasn't. So each of these people proved the haters wrong by winning... which meant that later, when running for their second terms with the haters saying don't do this, you will lose, their internal experience was "that's what you said last time and I won anyway, so I'm going to ignore you." (In a weird sense, the only people who get to a point of running for President without this particular kind of selection effect are nepo babies, of one kind of another.)
4) There's a meme in the US called "Florida Man", where you post a photo of a news report of a person getting arrested for doing something absolutely bananas ("robbed a bank while riding an alligator and using a banana") and inevitably they're from Florida. People use this to say Floridians are especially crazy. But actually, Florida just has a transparency law that makes it easy for news orgs to publish mug shots, so it's all selection bias.
I was told once that I lose readers by sticking overly-long overly-abstract preambles before my posts, so I'm sticking this one in the middle; if you don't like it, skip to the next section break.
Basically: eight out of ten of the last British prime ministers attended Oxford. Since only ~1% of British undergraduates attend Oxford, if Oxford acceptance were done by lottery and they still got these kinds of results, this would be a glowing endorsement of Oxford's educational interventions.
But of course, we all know that getting into Oxford is non-random: the kinds of people with the willingness and/or ability to become Prime Minister are unusually likely to go to Oxford in the first place, and if for some reason they hadn't gone to Oxford they would still be unusually likely to become Prime Minister. Heck, at this point, even choosing to apply to Oxford rather than Cambridge is statistically an indicator that you hope to become Prime Minister some day. (In the last hundred years, more Prime Ministers didn't go to university than went to Cambridge).
Which is to say there is a massive selection effect: individuals "select" into a certain group (either literally or metaphorically) based on pre-existing traits, which means that observed outcomes are potentially caused by these pre-existing differences rather than by the treatment itself.
One of my major beliefs about the world is that SO MUCH STUFF that people observe about the world is really just selection effects in different guises, such that one of the most useful things you can do is whenever you hear a claim about aggregates is to immediately ask yourself "how might this be a selection effect?"
[the following two are self-plagiarized from stuff I wrote previously, sorry]
5) you know those ads from insurance companies stating that People Who Switched To [OurCo] Frequently Save $X? This is a blatant selection effect: people who check out your insurance co and find out they're going to save $$$$ are unusually likely to switch.
6) the most relevant person to ask if you want to understand the problems at most companies doesn't work there. Companies do internal surveys of their employees to try to figure out what's going right/wrong. But the really smart thing would be to survey a) ex-employees who left because the org sucks, b) people you offered jobs to but who turned you down.
[and now the rest of these are plagiarized either from my friends or my own brain, who even knows at this point]
7) every single stat comparing College Educated Americans in 1960 and today are just selection biased out the wazoo. In 1960, 8% of Americans finished four-year college and now 40% do, so any kind of stat about e.g. the political leanings of college students then and now is talking about a completely non-comparable set of people.
8) everything you see online is filtered through a massive selection funnel for virality. Stuff doesn't reach you because it's true, it reaches you because it's viral. Statistically, almost-nobody is reading this sentence, because "9 selection effects" is not a good viral article title, and text blogposts are not a good viral medium.
9) if you looked at my recorded weight over time and averaged it, the weight you would get is significantly lower than my actual average weight over time. This is because at the times when I'm in good shape I'm much more motivated to consistently weigh myself every day, whereas at the times when I'm gaining weight I also stop weighing myself consistently.