[s9w]

The thing is that it's simply not exactly stated what kind of population they tested. Whether they were representative of that town or the German population. Of their sample was actually representative of the German population (which I heavily assume since they talked a lot about that) then the results will be quite robust.

[pintxo]

How would you create a representative sample for the whole population if all your samples are from a sub-population with a known cluster of cases?

[s9w]

The cluster thing is obviously hard to remove. But the population.. that should be very doable. You'll just have to adjust the people you chose according to the general population. Not sure exactly how that goes down in practice. But in the press conference he stressed multiple times how this was representative and was done in cooperation with statisticians etc.

[pintxo]

Problem here comes from the question "representative for what"?

I can easily see this to be representative for the town in question.

But I find it quite obvious that it cannot be representative for the whole country due to the initial difference in exposition.

[s9w]

Yeah well but is that really important? If they nail the same sample age-wise etc and use that to calculate lethality, then that number can be used on the general population, right? The only difference left is the rate of infection. But how many of those died should be the same.

[mannykannot]

I would guess that the spread of the disease and the rate of infection could be parameters to consider in planning policy for the immediate future. On the other hand, could this result be used to usefully estimate the total values in other regions from the hospitalization numbers?