right now I am stuck on problem. I need to find the sample size for an experiment with Hoeffdings inequality.
I want to know how many people I should sample so that the estimate differs from the true porption at most by 0.1 with a probability more than 99%.
I would solve this with Bernoulli Version of Hoeffding's Inequality.
If the question was asked as at least question I would solve it like that P(|x̄-p|>0.1) =< 2exp(-2nt^2) 0.99=< 2exp(-2n0.1^2) and so on.
My suggestion for solving at most would be (P(|x̄-p|<0.1)=1-P(|x̄-p|>0.1)>1-2exp(-2nt^2).
Is that the right way?
