[thebruce87m]

I can give you an example right now on something I’m working on.

I have an algorithm that either gets something correct or it doesn’t. My measurement system is also not infallible and will sometimes give out the wrong data, such that it is not possible to know whether the algorithm got it wrong, or the original data was wrong.

We still use the combined error as a KPI since it gives an overall health of the system, even though we cannot know the individual error rates. If the KPI tanks it tells us something is wrong and we can investigate. Could be the algorithm, could be the measurement equipment. It’s not perfect, but not having any KPI would mean we would be completely blind to any problems at all.

You will find the world is not perfect and compromises have to be made.

[pfannkuchen]

Is the output of your work used to inform the public on a health danger which determines their level of fear and support or opposition to response measures?

Obviously something like this is reasonable in some cases. Just because it is reasonable in some cases does not mean it is reasonable in every case.

Compromises should not be misleading, let alone extremely misleading.

[thebruce87m]

So what was the alternative? Hide the measurement? Not measure at all? What specifically would you have done differently?

Once the excess mortality data came along it confirmed people lots more people were dying than normal, and therefore of covid, so how is it misleading? If anything saying with covid is the opposite of misleading since you are specifying the caveat of the measurement up front?