Assume we got one out of range measurement out of 100. We can't say anything about variance of reading if we don't have any prior on variance ie using bayesian statistics. That's the joke that traditional statistician can't figure out the obvious thing.