That is much less likely than the other way around, though.
What the posterior over the expected post quality of the OP is, and what one should say in public about a stranger are two separate issues. Do you understand that if I start following your posts and replying to good ones effectively saying “boy, this was a good post, but it’s probably better than your average”, then that behavior would be extremely patronizing?
Bayes theorem is just a one liner consequence of the definition of conditional probability. Having social graces is actually hard.. and yet “being Bayesian” seems to be a bigger deal than being a good sport.
Now that this has turned into a discussion about statistics, it’s more important than politeness.
There are times when complaining about a social rule being broken is a worse infraction than breaking it in the first place. Your injunction to avoid pointing out flaws in someone’s reasoning because it’s impolite has left a far worse taste in my mouth than any insult by prior probability of the form: “you’re probably about average quality for a poster, because one post isn’t enough to prove otherwise.”
Now that this has turned into a discussion about statistics, it’s more important than politeness.
any insult by prior probability of the form: “you’re probably about average quality for a poster, because one
post isn’t enough to prove otherwise.”
What is insulting is what you are choosing to privilege. There are all sorts of things that are true about someone. For example, I am ethnically a Russian male. A fact about Russian males is that they have a very low average life expectancy (for obvious reasons). We could privilege this fact and (in a discussion about cryionics say) point out that, without additional evidence, I am likely far closer to death than an ethnic American my age. And because of this I should consider signing up for cryonics more seriously than an American my age.
This would be true (in fact not about me, because I almost never drink, but the “statistical reasoning” is sound), but an extremely socially stupid thing to say. Knowing what true things to privilege is the difference between Leonard and Sheldon on Big Bang Theory.
“Regression to the mean” as used above is basically using a technical term to call someone stupid. These sorts of “fact reporting” events don’t exist in isolation but in a larger social context, where people might use them to assert dominance and all sorts of other things.
I think your report that “this” is leaving a bad taste in your mouth is extremely fascinating to me.
That is much less likely than the other way around, though.
It’s the only post by this poster so far, so it’s exactly at the mean level of their posts. Going by that, we’d expect the next post to be of the same level as this one.
On the other hand, if you want to look at the mean level of other people’s posts, it’s not clear what reference class to use. All posts ever? All posts except the really bad one (because this poster doesn’t seem to be a troll / stupid / misguided)? All first posts? All first posts by people who have been reading LW for as long as User:gothgirl420666, however long that was? The posts by the same username on Reddit? Posts on subjects related to this one? Posts grouped with this one by some classifier algorithm?
That is much less likely than the other way around, though. That’s the point of regression to the mean.
But this is still evidence that his future posts will be at least somewhat above average! If not quite as good as this one.
What the posterior over the expected post quality of the OP is, and what one should say in public about a stranger are two separate issues. Do you understand that if I start following your posts and replying to good ones effectively saying “boy, this was a good post, but it’s probably better than your average”, then that behavior would be extremely patronizing?
Bayes theorem is just a one liner consequence of the definition of conditional probability. Having social graces is actually hard.. and yet “being Bayesian” seems to be a bigger deal than being a good sport.
Now that this has turned into a discussion about statistics, it’s more important than politeness.
There are times when complaining about a social rule being broken is a worse infraction than breaking it in the first place. Your injunction to avoid pointing out flaws in someone’s reasoning because it’s impolite has left a far worse taste in my mouth than any insult by prior probability of the form: “you’re probably about average quality for a poster, because one post isn’t enough to prove otherwise.”
What is insulting is what you are choosing to privilege. There are all sorts of things that are true about someone. For example, I am ethnically a Russian male. A fact about Russian males is that they have a very low average life expectancy (for obvious reasons). We could privilege this fact and (in a discussion about cryionics say) point out that, without additional evidence, I am likely far closer to death than an ethnic American my age. And because of this I should consider signing up for cryonics more seriously than an American my age.
This would be true (in fact not about me, because I almost never drink, but the “statistical reasoning” is sound), but an extremely socially stupid thing to say. Knowing what true things to privilege is the difference between Leonard and Sheldon on Big Bang Theory.
“Regression to the mean” as used above is basically using a technical term to call someone stupid. These sorts of “fact reporting” events don’t exist in isolation but in a larger social context, where people might use them to assert dominance and all sorts of other things.
I think your report that “this” is leaving a bad taste in your mouth is extremely fascinating to me.
Well I definitely wasn’t implying that. I actually wanted to discuss the statistics.
It’s the only post by this poster so far, so it’s exactly at the mean level of their posts. Going by that, we’d expect the next post to be of the same level as this one.
On the other hand, if you want to look at the mean level of other people’s posts, it’s not clear what reference class to use. All posts ever? All posts except the really bad one (because this poster doesn’t seem to be a troll / stupid / misguided)? All first posts? All first posts by people who have been reading LW for as long as User:gothgirl420666, however long that was? The posts by the same username on Reddit? Posts on subjects related to this one? Posts grouped with this one by some classifier algorithm?