286168_en-US

liliconcom

Sample Size calculation for certain CpK

Hi all,

 

a customer wants to know the optimal/minimal sampling rate to achieve a certain CpK-value.

 

They produce 150000 parts/year and require a Cpk = 1,67. Their question is:

How many random samples do we have to take with a batch size of 15,000 pieces per production unit.

 

Anybody have an advice? Is there a tool in JMP that can do such a calculation?

 

Thanks for any help:)

 

Best,

Florian

Re: Sample Size calculation for certain CpKHey! That’s not currently in JMP. This book (link below) has the calculations. They’re pretty straight forward: https://www.amazon.com/Sample-Size-Determination-Power-Thomas/dp/1118437608/ref=nodl_Re: Sample Size calculation for certain CpK

This would be a good item for the internal wish list.

 

Re: Sample Size calculation for certain CpK

@mia_stephens - I'll take the ball on getting in the internal list.  I also have an example of a deal where this was a sticking point for the potential customer as well as a couple existing customer examples where this was requested.  

 

Best,

 

M

Re: Sample Size calculation for certain CpK

I wonder if @calking has any thoughts or suggestions here.

Re: Sample Size calculation for certain CpK

I agree that adding a wishlist item would be useful. At the very least, it's something we can consider for the next version...

 

As for addressing the current concern, I see two possible approaches. One is to use the confidence interval formula at this link. You could easily create a data table with varying sample sizes in one column and the confidence interval bounds in the next two columns (maybe using the target Cpk as the estimate in the formula). The customers can then determine the appropriate sample size based on a desired interval size. 

 

Another approach is to create a column of simulated CpKs using a formula, where you can go in and change the sample size and other parameters. You can then have a table script that runs Distribution on the results to see the variability in the estimated CpK. You can then determine sample size based on the amount of variability you want in the estimated CpK. I've attached a simple data table with an example. This is very similar to the confidence interval approach, so you might go that route if its simpler. 

Re: Sample Size calculation for certain CpK

Hi Caleb,

 

is it ok if I publish your answer as a community post?  So that we can forward it to customers?

 

Thanks

FLorian

Re: Sample Size calculation for certain CpK

Florian, I believe Caleb is out on paternity leave. I think it is OK to post his answer on the community.

 

Mia

Re: Sample Size calculation for certain CpK

Perhaps the best way to do this would be to ask the customer to post the question in the Discussions forum and then you can post your answer in response.