Statistics expert - cumulative binomial stuff

[roscoe]

My job requires me to establish the appropriate sample size during very expensive flight testing (i.e. can't afford to do more testing than the minimum required to establish the appropriate confidence). I have a nomograph that was given to me that we use to calculate the right value given confidence and demonstrated reliability (i.e. results). The nomograph can be found at: https://acc.dau.mil/docs/dtepi/pns/doc/nomogr/bdi.pdf.

Assuming we want to show that we have 80% confidence (right axis) that we'll succeed 80% of the time (left axis, also called reliability), I connect those values on each side with a straight edge and the chart shows that with zero failures (upper right boundary) we need 7 trials. If we have one failure (upper left boundary) the chart shows that we need 12 trials (11 successes). I was provided a formula that defines the zero failure curve [n=Log (1-confidence)/Log (reliability)]...What I need is a more generic equation or excel function that will generate these curves with failures not equal to zero.

Any statistics experts out there that can help me?

Thanks

 

[roscoe]

Nobody?

 

[c_m]

Roscoe,

Your problem is kind of tough. However, I have some links that you might be interested in and that may help you solve this problem. See:

http://en.wikipedia.org/wiki/Binomial_distribution
http://www.itl.nist.gov/div898/handbook/prc/section2/prc241.htm

I will post if I come up with something.

Btw, is there any other information that you can provide besides what you have already posted?

 

[roscoe]

Nope. We've been doing this for years because it's "the way we've always done it". I hate that and want to be able to support the numbers.

As for the pages you reference, those are over my head (I've forwarded them to the prob/stats expert in the office.