Im trying to get my head around the relationship between the PPK for continuous data and the ability to make a Confidence and reliability statement when using a certain amount of samples to perform your testing. As an example if you wanted to make a 95%/99% CR statement and you observed that a Pre-PQ observation was a PPK of 1.09, then as per a table this allows you to choose 80 samples and as long as you get a PPK of>= 0.92 you can then make 95/99 claim. The table gives several examples using a Pre_PQ observed value so you pick the One in the table which is closest to your observed value and lower (1.05 is close and lower than 1.09).

Ive being scouring the web to find an explanation of this but have had no luck. Can any of you point me in the right direction.?

Thank you.

Pre-PQ observed Ppk (95% confidence lower bound) | Sample Size | Acceptance Criteria | LTPD0.05 |

Ppk≥1.13 | n=50 | Ppk≥ 0.96, Pp≥ 1.02 | ≤ 1% |

Ppk≥1.11 | n=60 | Ppk≥ 0.95, Pp≥ 1.01 | ≤ 1% |

Ppk≥1.05 | n=80 | Ppk≥ 0.92, Pp≥ 0.99 | ≤ 1% |

Ppk≥1.02) | n=100 | Ppk≥ 0.90, Pp≥ 0.97 | ≤ 1% |

What does "Pre-PQ" mean? What does LTPD0.05 mean? I can guess, but best to know what the author of this table used. And what is the source of this table? That would help in answering the question.

The table does seem to behave similar to the Binomial. At a high Ppk (previous low probability of defect) you need less samples to show that the chance of failure with the new population is good. Similarly, with a low Ppk (previous high probability of defect) you need much larger sample sizes to show that the new population is good. This relationship could probably also be proven with Bayesian methods.

But again, we are dealing with a third order metric - taking an actual measurement, converting it to go no-go, and then manipulating it with the normal distribution. When it may not be normal. So I am with Bev in saying this is voodoo statistics, where we actually should be working with the real data and/or creating simulations to evaluate the sample sizes needed. Even better, use Statistical Process Control, again going be to the ground "truth" rather than manipulated data.

By the way, been traveling - my wife and I just got back from Japan.