Hacker News Reader

Comment by huac | Hacker News Reader

> Most of his recruits speak a language familiar to engineers but foreign to anybody else: Six Sigma, Kaizen, Lean. Woodburn is prone to fall into such jargon. “Anyone who’s been to an airport can say that it’s operating at maybe Two Sigma,” he marvels. “That’s making a mistake 30 percent of the time.”

Two sigma corresponds with a 95% confidence interval, while one sigma is a 68% confidence interval which he is presumably referring to.

I don't want to detract too much from the work that these people did, which I think is actually very cool, but it worries me that their head of engineering can't discuss really simple statistics correctly...

Edit: words mean nothing, see replies

View on HN

For whatever reason, the "Six Sigma" techniques/ideas aren't using "Two Sigma" to refer to ±2σ, instead they're referring to the width of the band around the mean.

https://en.wikipedia.org/wiki/Six_Sigma#Sigma_levels

> For whatever reason, ... they're referring to the width of the band around the mean.

That's exactly how I learnt it in secondary school, the standard deviation, what's the problem?

Yeah, and it's apparently also shifted down by 1.5 sigma.

That's apparently based on Motorola's data regarding process drift.

> Motorola has determined, through years of process and data collection, that processes vary and drift over time – what they call the Long-Term Dynamic Mean Variation. This variation typically falls between 1.4 and 1.6.

https://www.isixsigma.com/new-to-six-sigma/dmaic/15-sigma-pr...

2 sigma is a defective rate of 31%. [0]

[0] https://en.wikipedia.org/wiki/Six_Sigma#Sigma_levels

I'd also like to point out (aside from the already clarified term 2 Sigma), mistakes like that are more likely the fault of the journalist and not the engineer.

Also, doesn't the +/- sigma rule only apply to a normal (gaussian) distributions?

Yes, that rule:

   31%  outside +/- 1 standard deviation  ("2 sigma"), 
   5%   outside +/- 2 standard deviations ("4 sigma"),
   0.3% outside +/- 3 standard deviations ("6 sigma"),

etc. However, the more general Chebyshev inequality states that for any distribution, you have at most 1/k^2 outside +/- k standard deviations, so (at most):

   100% outside +/- 1 standard deviation, 
    25% outside +/- 2 standard deviations,
     9% outside +/- 3 standard deviations,

etc.

https://en.wikipedia.org/wiki/Chebyshev%27s_inequality

Reply on news.ycombinator.com