Sample Size : MatchedPairs + Indexing to an AQLHi,
A user asked me the following question. They would like to determine the sample size in the following 2 cases:
- Sample sizes to be determined in the matched pairs case (comparison of the performance of 2 diagnostic tests evaluated on the same population, where the correlation between the results of the two tests is known to be non-zero),
- Evaluation of sampling plans indexed to an AQL (for both attribute testing and measurement-based testing in accordance with ISO 3951-1), i.e. using JMP to calculate the PRQ and CRQ of the plans and access the efficiency curve for each plan.
AQL: Acceptable Quality Level, PRQ: Producer's Risk Quality, CRQ: Consumer's Risk Quality
Could you help for both cases?
Re: Sample Size : MatchedPairs + Indexing to an AQLRegarding the first request, the Matched Pairs platform performs a basic one-sample t-test of the pairwise differences so that you can use a simple size calculator or explorer for one sample.mean. Matched Pairs further provides some nice diagnostics for data issues or the assumptions of this test. JMP 17 also added the Bland-Altman analysis for method comparison.
Re: Sample Size : MatchedPairs + Indexing to an AQLFor the 2nd request, the Quality and Process, OC Curves platform will do AQL calculations for single and double inspections. Fair warning: when people see how high the probability is of accepting bad product from sampling inspection, they usually adopt other methods unless they are forced into sampling by things beyond their control (Wheeler's Green EMP book has a fun chapter explaining this).
In the example below, there's a 5% chance of accepting a batch of product that is 10% defective, based on sampling 25% of the product (25% sampling is absurdly high in my experience).
Re: Sample Size : MatchedPairs + Indexing to an AQLThanks a lot Jed! It really helped me. I have to admit, I'd never opened this platform before today.
