Spehro Pefhany wrote:

> They said "The failure rate could be as high as 30%.". It "could" also be
> as low as zero, perhaps.

They also didn't say 30% of what... I'm always surprised how much
discussion can happen about the actual value of a percentage without
anybody knowing the base of that percentage.

If I buy one part and it's defective, my failure rate of that part is 100%
(at least when using the "popular" meaning of % -- keep in mind that we're
not really talking about statistics here :). The failure rate of the lot
this part was in probably was something else, the failure rate of the waver
batch, of the whole foundry, of the packaging machine, of the manufacturer
overall, ... you get the idea. All different average failure rate
percentages.

So what's the reference of the 30%? The truck load? Give me a break. I
don't think they have serious statistics about defect averages in truck
loads... :)

Then the "could be as high as"... Employing statistics, for any finite
selection with finite error rates you get a non-0 probability that there is
a failure rate of 100%. And you get a non-0 probability that there is a
failure rate up to 30%. So in precise statistical terms, they didn't say
squat and just stated the obvious -- yes, the error rate could be as high
as 30%, which is something that can be said to have a probability of
greater than 0 for any selection of components. To become real, non-obvious
information, they would have to have included some more details about their
statistic assumptions.

Gerhard
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist