3 ms·
> there isn't a hard theorem why they should sit at +/-3σ There is Vysochanskij-Petunin inequality [1]. 95% of observations should fall within +/-3σ away from
by pps43 6y ago
> there isn't a hard theorem why they should sit at +/-3σ
There is Vysochanskij-Petunin inequality [1]. 95% of observations should fall within +/-3σ away from the mean, assuming unimodal distribution and finite variance.
[1] https://en.wikipedia.org/wiki/Vysochanskij%E2%80%93Petunin_inequality https://en.wikipedia.org/wiki/Vysochanskij%E2%80%93Petunin_i...
- conformist 6y agoThanks, I wasn't aware of this and it seems very useful! Good to know a tighter bound than Chabyshev's inequality (assuming unimodality).
- oso2k 6y agoIt's also where Six Sigma (3 sigma on each side) comes from. It implies "Total Control".