Then that’s not such a bad situation. For a log-normal (or extremal) distro even a single extremal observation will move the mean by a lot, so it’s not so bad to be overshooting what your current data shows you…
This depends on your use case however.
1 Like