Re:33 Questions
by thethimble on Sunday November 24, 2013 @04:17PM
(#6791169)
All questions must be independent. That's too hard. I would be surprised to find even a handful of questions that have no statistical correlation when applied to all people in the world.
Re:33 Questions
by enriquepablo on Sunday November 24, 2013 @04:40PM
(#6791254)
> To properly mimic this property with yes/no questions, you will have to come up with questions that divide the whole Earth's population equally AT EVERY NEW QUESTION.
Not only that: The first question must divide the population in 2; the second question must divide each of the 2 subsets produced by the first question in 2; and the third has to perfectly divide the 4 subsets produced by the second... And this independently of the order in which you make the questions.
To properly mimic this property with yes/no questions, you will have to come up with questions that divide the whole Earth's population equally AT EVERY NEW QUESTION. Even the most obvious one, "are you (fe)male?" is slightly biased toward men (according to wikipedia). At every question that skew your 50/50, you'll have to add another question beyond 33 to catch up with this.