On Mon, May 13, 2013 at 5:56 PM, Jed Rothwell <jedrothw...@gmail.com> wrote:
> Kevin O'Malley <kevmol...@gmail.com> wrote: > >> >> ***We can proceed with the same probability math I used upthread. If one >> considers it to be 1/3 chance of generating a false-positive excess heat >> event, then you take that 1/3 to the power of how many replications are on >> record. >> > > That is a form of Bayesian analysis, I think. > > No it's not. It's just ordinary probability theory, and it's not even right. That calculation gives the probability of getting N *consecutive* replications. The probability of rolling 6 on an ordinary die is 1/6, but it's easy to get N sixes (on average) just by throwing the die 6N times. It is the need for these sorts of arguments and Bayesian analysis that emphasizes the absence of a single experiment that will give an expected result.