[majos]

Confused — the article starts off with the reasonable point that we don’t know the true number of covid infections (likely only the most severe ones) so calculations of death rates are flawed. But then they address this by instead computing # deaths that day/# infected 14 days prior? How does this solve the initial problem?

[rolph]

they are attempting to relate the infections to an average outcome 14 days later, this seems to view the dynamics as a series of interleaved 14 day cycles.

in this case if one has a number of outcomes that resolve in death or recovery and that number is larger than your original number for infections, then you can fudge the numbers after and say thats the number of infections we didnt know about 14 days ago in order to adjust.

this is interpolation and if it occurs it must be made obvious in your data presentation. it really does help to provide an expected value to compare to a future observed value. youll be able to correct and get in the right ball park, then ask where the deficiencies in theory are that provide explanation for these differences.

[thedudeabides5]

Good point.

If you assume like 50-75% underreporting, the mortality rate get cut in half/by a quarter