3 ms·

> Adding this addendum makes it pretty clear that this data is far from generalizable in any meaningful way. Why? We have real world data and can use inductive

> Adding this addendum makes it pretty clear that this data is far from generalizable in any meaningful way. Why? We have real world data and can use inductive reasoning to estimate what the "typical woman" thinks of the "typical man." You're right to point out that this has flaws, but that doesn't mean it's not meaningful. If nothing else, it's a far better estimate than just guessing. That is, unless you have some other evidence that points in another direction.

Etheryte 2y ago
A narrow self selected subset of the population, people who use dating sites and apps, doesn't tell you anything about the population as a whole. The selection is not uniform, you could observe literally any result on the wider population. Imagine you have a banking app for the ultra rich. Would you be able to infer anything about the general population from the spending habits of your users? No, because they're not indicative of the population as a whole. The case is the same here.
- bitshiftfaced 2y ago
  In your example, there's presumably a strong correlation between the users' (and their wealth) and their spending habits. That said, we probably could infer some things about the general population from that group. For one, they are part of the population. Two, there will be some similarities between their group and the complement of their group because humans have similar needs (e.g. groceries, gas, etc.). If you took this as your prior for the supergroup, then you would be more accurate than if you took a random number. There are no doubt differences between women on Okcupid and women in the general dating population. But in your example, we have a clear reason for believing that wealth is correlated with spending habits. If there is a reason why you believe Okcupid users should be as correlated with poor evaluation of the typical man, then it is at least not clear to me. Once again, using this data is likely far better than taking a random guess at how women in general evaluate the typical man.

Etheryte 2y ago

A narrow self selected subset of the population, people who use dating sites and apps, doesn't tell you anything about the population as a whole. The selection is not uniform, you could observe literally any result on the wider population. Imagine you have a banking app for the ultra rich. Would you be able to infer anything about the general population from the spending habits of your users? No, because they're not indicative of the population as a whole. The case is the same here.

bitshiftfaced 2y ago
In your example, there's presumably a strong correlation between the users' (and their wealth) and their spending habits. That said, we probably could infer some things about the general population from that group. For one, they are part of the population. Two, there will be some similarities between their group and the complement of their group because humans have similar needs (e.g. groceries, gas, etc.). If you took this as your prior for the supergroup, then you would be more accurate than if you took a random number. There are no doubt differences between women on Okcupid and women in the general dating population. But in your example, we have a clear reason for believing that wealth is correlated with spending habits. If there is a reason why you believe Okcupid users should be as correlated with poor evaluation of the typical man, then it is at least not clear to me. Once again, using this data is likely far better than taking a random guess at how women in general evaluate the typical man.