Simple stats problem: number of tests to be significant?

[chaoticfarmin]

Hey guys,

Basically I have a component that I want to test multiple times. The outcome of each test is simply a Pass/Fail. What I want to know is how many tests do I have to do to be (say) 95% sure that I have the right result?

My stats knowledge is a bit hazy but I know that for these type of yes/no situations it may be possible to use the Binomial distribution. Am I on the right tracks?

Cheers if you can help.

[Bacle2]

This is an idea; I don't know what the real-life constraints you are working with, so here is a general idea: you can use the CLT to estimate the mean: take many random, independent samples (tests, with results of pass/fail, or 1/0) of the same size N: for each sample compute the mean number of sucesses (total successes/total trials). By the CLT, this random variable, the sampling mean, has a normal distribution. Then use a 95% confidence interval about the samplingn mean.