Sanity Check on AQLs (Acceptable Quality Level)

[plastic gal]

It has been a long time since I have used attributes sampling and I want to see if I still can interpret the table. If I choose an AQL and keep the Ac = 0 & Re = 1 fixed (in other words don't allow any rejects) and the arrow points down to where say the Ac = 1 and Re = 2, do I have to use the sample size associate with the later Ac/Re # (which is a larger sample)? I believe that I do and just keep the Ac = 0 and Re = 1. I am not a big fan of attributes sampling but it is a customer requirement.

Am I way off-base?

Thanks........................

[Geoff Withnell]

Originally Posted by plastic gal

It has been a long time since I have used attributes sampling and I want to see if I still can interpret the table. If I choose an AQL and keep the Ac = 0 & Re = 1 fixed (in other words don't allow any rejects) and the arrow points down to where say the Ac = 1 and Re = 2, do I have to use the sample size associate with the later Ac/Re # (which is a larger sample)? I believe that I do and just keep the Ac = 0 and Re = 1. I am not a big fan of attributes sampling but it is a customer requirement.

Am I way off-base?

Thanks........................

plastic gal,
If you follow the process as you describe, you will actually be checking to a lower AQL then the one you chose. If you prefer c=0 plans (I do for a number of reasons), may I recommend
"Zero Acceptance Number Sampling Plans", Fourth Edition
Squeglia, Nicholas L.
It is available from ASQ Bookstore, and probably Amazon, and has c=0 tables equivalent to the ANSI or MIL 105 tables you are probably using.
I have used it with good customer acceptance and good results.

Geoff Withnell

[plastic gal]

Thanks so much for the reply. That is what I figured and it is nice to have the confirmation as such. I have also used the c = 0 plan in past lives.

So what I am hearing you say is that you must follow the arrows to the larger sample size no matter what. In this case it is for a destructive test for rockwell hardness (have to remove oxide layer) so I don't want a huge sample but too risky to have rejected product.

Thanks again - if there is anything else you can share it is much appreciated. Dr. Deming said that acceptance sampling guarantees that the customer will receive defective product which is possible with any sampling that is not SPC based with a large Cpk.

Pamela

[Statistical Steven]

Originally Posted by plastic gal

Thanks so much for the reply. That is what I figured and it is nice to have the confirmation as such. I have also used the c = 0 plan in past lives.

So what I am hearing you say is that you must follow the arrows to the larger sample size no matter what. In this case it is for a destructive test for rockwell hardness (have to remove oxide layer) so I don't want a huge sample but too risky to have rejected product.

Thanks again - if there is anything else you can share it is much appreciated. Dr. Deming said that acceptance sampling guarantees that the customer will receive defective product which is possible with any sampling that is not SPC based with a large Cpk.

Pamela

Are you required to use the Z1.4 standard? If not, you can use the binomial directly and calculate the AQL and LTPD for ANY sample size and Ac/Re number. I find this method to be superior to the standard since it gives smaller sample sizes and gives me BOTH the consumer and producer risks which should be stated for ANY sampling plan.