tags: #publish
links: [[SRE and DevOps]],
created: 2020-10-21 Wed
---
# Average of averages problem
https://math.stackexchange.com/questions/95909/why-is-an-average-of-an-average-usually-incorrect
Averaging averages *without taking into account the number of datapoints in each average* gives misleading or meaningless results.
Intuitively, the problem is loss of the denominator: the relative contribution of large numbers of results close to the average gets lost when you average the averages.
Solution is Histograms: keep the frequency component of the data instead of collapsing into Averages, and only sum when analyzing.