-What we observe about the variance of mortality of corona seems about 0.02 or so, which is too big considered that the rates are obtained based on hundreds and thousands of people (like state level).
-So, the true death rate is hyper-parameter that we may assume that it has a beta distribution. Thing is that, when the death rate is constant for each state (not every individual has different death rate), the variance is bigger.
-When specificity is high, the variance is a function of sample size (independent of death rate).
-But, when specificity is low, the variance is not decreased with increasing sample size.
- On the other hand, high or low sensitivity does not change much of the variance (independent of death rate). large sample size indeed decreases variance.
-The real death rate based on the generated data is indeed function of death rate and sample size.
- So, the low specificity is the problem. What is the meaning of the low specificity in death?
A lot of people never get tested - who are mostly healthy. So this whole "system" of detecting "death" has very high specificity because most of them are considered healthy...
???
