3 ms·
These types of things can be guesstimated on validity pretty well, using likelihood. Do you know 10 people? 20? Ask how many have been swatted. If the answer is
by godelski 3y ago
These types of things can be guesstimated on validity pretty well, using likelihood. Do you know 10 people? 20? Ask how many have been swatted. If the answer is "none", then it decreases the likelihood that the model is accurate. The further off your sampling is from the "reported" rates, the more likely that that model is wrong.
The other explanation is that the data is heterogeneous but then that means we have an aggregation error. That also isn't great either, especially since it means the problem is potentially even larger, but only affecting a specific group.
Either way, not good.