Thread:

The purpose of testing in an epidemic is not to find the number of people infected but find as many infected people as you can and swiftly isolate them from the population to prevent further transmission

This is different from say an exit poll whose purpose is to accurately estimate an election winner or a market study to gauge consumer preferences, say for cellphones.

Purpose is everything that decides a sampling scheme

In an epidemic like covid19, I repeat, the purpose is to swiftly find as many infected individuals as possible and isolate (and treat) them to prevent further spread.

A sampling scheme needs to indeed be biased and look for high probability "suspects"

Here the sample size N is constrained by test kit availability. But irrespective of sampling scheme, test accuracy (false+ve/-ve) etc, one thing is obvious:

Expected no. of infected people discovered and hence isolated increases as N, ie sample size increases

So it is imperative that more and more testing be done to isolated infected people further. The sampling scheme, ie whom to test is a separate question independent of sample size

In some cases, false positives have deadly consequences. An innocent undertrial may be executed or an uninfected limb may be amputated.

In this case a false positive just means the person ia quarantined. Conditions may be bad and should be improved. But not disastrous like above

Conclusion. Test more (as much as you can). Test swiftly. Isolate and treat the positives.

This is irrespective of how you choose whom to test.

Regarding the sample allocation itself, here is an important consideration https://mobile.twitter.com/bhalomanush/status/1250427133727391746

Basically remember that your sample is very quickly time varying.

Also remember that unlike most other sampling, your sampling scheme's purpose is to change the dynamics of this sample (for the better)

As an analogy. If you run a restaurant,it would he extremely unethical for you to remove the 1* reviews on yelp or zomato, lest other ppl see it.

But in this pandemic, that is exactly what you must do. Removing the positives from the sample on war footing