Tickerguy
210k posts, incept 2007-06-26
2017-04-03 09:10:02
One of the first things you learn in Stats class:
An "error" is normally-distributed. Over a large enough sample size (n > 100, typically) an actual error will trend toward zero in aggregate, and very reliably will be within 1SD. Over even larger sample sizes this becomes even more-true. Confidence range shrinks as sample size rises, in other words.
An "error" that always goes one way, especially when the sample size is huge (e.g. n > 10,000) is not an error -- it is fraud.
----------
"Perhaps you can keep things together and advance playing DIE games.
Or perhaps the truth is that white men w/IQs >= 115 or so built all of it and without us it will collapse."
RSS available