How to Calculate the producer Risk

S

senthilgaurav

Again as i told you yesterday I use the MINITAB which can generate graphs and solutions once given all the valid inputs

Kindly advise the steps to key in values as I did download Minitab yesterday after saw your photo. So far i did not know about Minitab.
 

eule del ayre

Involved In Discussions
Steps for Acceptance Sampling by attributes
1. Click stat on the upper side of the options
2. Go to Quality Tools
3. Then proceed to Acceptance Sampling by attributes
4. Once you're on the Acceptance Sampling by Attributes
- on the upper side there's an option between: Create Sampling Plan and Compare User Defined Sampling Plans; Pick the Compare User Defined Sampling Plans.
5. On the Measurement type pick Go/ no go (defective)
6. On the Units for quality levels pick Percent Defective
7. On the Acceptable Quality Level (AQL): input .65
8. On the Rejectable quality level (RQL or LTPD): input 5 (you didn't include the RQL so i input 5)
9. Sample Size: 77
10. Acceptance Numbers: 0 (like you said the acceptance value is zero)
11. Lot Size: 30000 (as you said over and over)
12. Click OK after you have entered all your inputs
Please see attached file
 

Attachments

  • Acceptance Sampling by attributes.xlsx
    414.8 KB · Views: 637
S

senthilgaurav

Steps for Acceptance Sampling by attributes
1. Click stat on the upper side of the options
2. Go to Quality Tools
3. Then proceed to Acceptance Sampling by attributes
4. Once you're on the Acceptance Sampling by Attributes
- on the upper side there's an option between: Create Sampling Plan and Compare User Defined Sampling Plans; Pick the Compare User Defined Sampling Plans.
5. On the Measurement type pick Go/ no go (defective)
6. On the Units for quality levels pick Percent Defective
7. On the Acceptable Quality Level (AQL): input .65
8. On the Rejectable quality level (RQL or LTPD): input 5 (you didn't include the RQL so i input 5)
9. Sample Size: 77
10. Acceptance Numbers: 0 (like you said the acceptance value is zero)
11. Lot Size: 30000 (as you said over and over)
12. Click OK after you have entered all your inputs
Please see attached file

Thanks a lot. I did the same and got the same result same as you.
Still, want to clarify that we put as RQL 5% is the maximum allowable percentage of Rejection for supplier??? I don't have any idea of it so far
Then the 1.282 % means, 385 defective pieces are allowed for the lot qty 30000/ if customer agreed for RQL 5% they supposed to be accepted the lot even it is with 385 pcs of defective parts??? is it my understanding is correct??

Maybe my question will be crazy but I should answer to my customer by tomorrow until they convince.....that's y I have to be clear/ready for all of this questions. May be they will ask the question with me.
 

Bev D

Heretical Statistician
Leader
Super Moderator
Again there is no. Single number of defectives that might be in the lot when you have found zero defects in your sample. I know you want this answer but it isn’t possible. The answer is the OC curves. There is always some probability that there are defects in the lot. The lower the actual defect rate the higher the probability that you will not detect it in your sample. The higher the defect rate the lower the probability that you will not detect it. For example, your plan allows a 5% probability that you will accept a lot with 3.8% defects in it.

How many defects did your Customer find? What quality level does your Customer require?
 
S

senthilgaurav

Again there is no. Single number of defectives that might be in the lot when you have found zero defects in your sample. I know you want this answer but it isn’t possible. The answer is the OC curves. There is always some probability that there are defects in the lot. The lower the actual defect rate the higher the probability that you will not detect it in your sample. The higher the defect rate the lower the probability that you will not detect it. For example, your plan allows a 5% probability that you will accept a lot with 3.8% defects in it.

How many defects did your Customer find? What quality level does your Customer require?

Thanks a lot to keep responding to for my inquiries so far...
The problem is our customer is using ANSI/ASQ Z1.4, AQL 0.65, so the sampling qty is 300 (Level II & M ) and found 40 pcs with faulty.
lot qty 30000
AQL 0.65
C=5
5=6
I also agree that they reserved the right to reject per ANSI/ASQ Z1.4 STD. Because if I am not wrong per AQL 0.65% with more than 1.6% of rejected part (Round 2%) the lot will be rejected.

As they also not familiar with C=0 sampling plan same like me .....for them no matter what sampling plan we are following but they want to the percentage of defect we are allowing in full lot qty per AQL 0.65, C=0 Sampling plan that's all.

As i cannot get the allowable defective for each and every sampling qty from the table of c=0 sampling plan, for me unable to answer to them, that's the reason I trying here to get an answer for them at the same time to get understand myself.
 

Bev D

Heretical Statistician
Leader
Super Moderator
well c=0 plans are not that great. people think they are good because they allow for no defects in the sample, but the fact is that smaller sample sizes are just not that great at detecting low defect rates. Unfortunately people don't fully understand the probabilities involved in sampling plans.

Your Customer's plan is better than yours and would detect defect rates higher than .65% much more frequently than yours.

For example, if they found 40 defects in their sample of 300 and that defect rate is representative of the lot, then your plan should have also caught this defect rate most of the time (78% probability of catching it). Although a 22% chance of missing it isn't all that small. see how the probabilities work?

Of course the thing that can't be captured in the math is the potential that the defects weren't homogenously distributed in the lot and/or you didn't follow a truly random sample plan. it simply isn't possible to quantify those probabilities ahead of time.

Sampling plans aren't very good at 'guaranteeing' quality. that has to be built into the process.

this situation will come down to negotiation with your Customer. if neither of you are very knowledgable about probability it will be a difficult discussion. Sorry I can't help you more but you really need to have a deep understanding of sampling plans and their limitations - or hire someone who does...
 
Top Bottom