Verification Sampling vs. AQL - Sampl Sizes

[kennethshaw]

Hi,

I've been tasked with training our group on sample sizes for verification testing. The group has historically used AQL to come up with minimum sample sizes, but I've relied more on confidence intervals and power analysis.

Can anyone provide guidance?

Thanks.

Ken

 

[Bev D]

there are several threads here addressing this very point. try searching fro sample size and verification or validation.


to the point: AQL based sample sizes are for acceptance sampling of a single lot or batch. they are intended to NOT detect any defects if the defect rate is at or less than the acceptable defect rate (AQL). they are also categorical data based...and so have very little resolution when continuous data is involved/available. they have very little 'power' when it comes to verification or validation testing which is when you actually want it.

confidence intervals and general sample size formulas (using alpha/beta risk and delta) are for single and multi-factor experiments, study designs & analyses (DoE/hypothesis testing) and verification & validation studies. There are formulas for every data type, level of precision and risk.

SPC and MSA (aka GR&R) require even different sample size considerations...

the formulas for them are different because they have different intents. your approach is correct.