2013-07-22, 11:47 AM
hadriel Wrote:As an idiot in statistics, I'd do a very rough estimate for a minimum sample size:
For a 5% chance of something happening, to have an average of 95% probability of that something to happen, you need something like 60 cubes. But a single instance of that happening is not statistically strong - you need like... up to 10 times that number. 600 cubes is... anyone wants to shell that kind of money for the sake of spading NexonAM's numbers?
Hadriel
One rule a lot of people like to use is np > 10 and nq > 10. p = probability of success, q = probability of failure. So at least 10 successes and 10 failures to almost guarantee accurate estimates and tests on the true value of p.
That's still ~200 cubes (10 success 190 fail) when p = 5% though. It gets pretty bad when p is close to 0% or 100%.

