Read on Twitter
Heads up econofolks: data that are *generated* by testing -- such as, say, indicators of who has a disease or who has antibodies in their blood -- are subject to various false positive / false negatives issues that the data we use in economics usually are not.
Also, if the goal is to infer rate of X in the population, non-random sampling from that population (especially with selection in based on unobservables) is going to absolutely torpedo that inference in a way it *might* not have for inferring relationship between X and Y say.
which is all to say, the toolbox and set of basic assumptions you're used to in econometrics is NOT a useful guide to this problem. PLEASE STAY HOME INTELLECTUALLY AS WELL AS PHYSICALLY.
