Dave - yes there are some segments of industry that have laws - I too have some parts of our product line under FDA control and I've worked in aerospace and defense. (ugh!)
However, this guy was doing what amounts to a 'dock audit' after a 100% inspection. And given the other things he could do that I listed - he could fully justify statistically, scientifically and even logically a relatively small sample that wasn't derived from traditional statistical plan approaches such as an AQL (an outdated useless but extrememly confusing system by the way) which was my meaning although sloppily articulated.
A traditional approach such as the AQL system is based on the assumption and intent that one is only sampling. And therefore one has to specify some defect level that one wants to have some specified confidence in detecting. If you really are looking to protect yourself from very low defect rates you will get a very large sample size. BUT - and this is the key but - he isnt only sampling he is doing is doign a sample after a 100% inspection. If he were to use the AQL he would have an uneccessarily large sample size.
There is a big difference between justifying a particular sample plan with statistics and using a statistically based sample plan generator like an AQL system: hence your own bolding: valid statistical rationale.
even if he were in a highly regulated legal environement if he wre to do the things I listed he woudl have a valid statistical rationale.
However, this guy was doing what amounts to a 'dock audit' after a 100% inspection. And given the other things he could do that I listed - he could fully justify statistically, scientifically and even logically a relatively small sample that wasn't derived from traditional statistical plan approaches such as an AQL (an outdated useless but extrememly confusing system by the way) which was my meaning although sloppily articulated.
A traditional approach such as the AQL system is based on the assumption and intent that one is only sampling. And therefore one has to specify some defect level that one wants to have some specified confidence in detecting. If you really are looking to protect yourself from very low defect rates you will get a very large sample size. BUT - and this is the key but - he isnt only sampling he is doing is doign a sample after a 100% inspection. If he were to use the AQL he would have an uneccessarily large sample size.
There is a big difference between justifying a particular sample plan with statistics and using a statistically based sample plan generator like an AQL system: hence your own bolding: valid statistical rationale.
even if he were in a highly regulated legal environement if he wre to do the things I listed he woudl have a valid statistical rationale.
SA: