When to use Cpk or Ppk for a Process Capability Study?

Q

qakid

I am analyzing a process capability study and am not sure which capability index applies. For our thirty piece sample, I am looking at the Ppk because the subgroup is 1. Since the subgroup is one, I want to use the actual standard deviation.

Our other study is made up of thirty subgroups. The subgroup size is 10. For this study I am looking and the Cpk because it uses an estimated standard deviation.

Am I correct in how I am using the capability indexes?
 

Stijloor

Leader
Super Moderator
Re: Cpk or Ppk?

I am analyzing a process capability study and am not sure which capability index applies. For our thirty piece sample, I am looking at the Ppk because the subgroup is 1. Since the subgroup is one, I want to use the actual standard deviation.

Our other study is made up of thirty subgroups. The subgroup size is 10. For this study I am looking and the Cpk because it uses an estimated standard deviation.

Am I correct in how I am using the capability indexes?

Look here.

Stijloor.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Cpk or Ppk?

First of all, before applying any math to the issue, how were the samples taken? What is the distribution of the data? Do you have a run chart to determine if the output variation is truly random? How many factors of your "total variance equation" are influencing the measurement you are observing? Can you say it is just one is statistically significant, and therefore not multi-modal? Is it non-normal...or should it be? These questions have a lot more to do with the validity of Cpk or Ppk than the subgroup size.

You may want to attach your data...it helps us respond more specifically to your issues.
 
E

Ehsan Heidari

I use Ppk for initial samples of tools or for the first time of production and for the others I use Cpk.
 

Jim Wynne

Leader
Admin
I am analyzing a process capability study and am not sure which capability index applies. For our thirty piece sample, I am looking at the Ppk because the subgroup is 1. Since the subgroup is one, I want to use the actual standard deviation.

Our other study is made up of thirty subgroups. The subgroup size is 10. For this study I am looking and the Cpk because it uses an estimated standard deviation.

Am I correct in how I am using the capability indexes?

Aside from all of the relevant questions bobdoering brings up, the whole idea of Cpk and Ppk is vastly overrated and almost as vastly misguided. The whole idea behind sampling should be to sample such that the difference between the population standard deviation and the sample standard deviation is negligible.
 

Bev D

Heretical Statistician
Leader
Super Moderator
I'll add to Bob and Jim's comments:

Your OK using Ppk for the 30 piece sample (although how the samples were taken - random and representative of the process) is FAR more important than which formula you use.

For your 30 sugroups of size 10, whether you use the Cpk formula or the Ppk formula it is important to understand if your process is homogenous. If it isn't Cpk will understimate your true performance. However, if you are only interested in the potential future capability then use Cpk.

Basically - even tho the formulas are fairly straightforward there is a whole lot to understand about capability indices and even more to understand process capability.

As Bob suggested - if your serious - please post your data, tell us about the process and what you are trying to know about the process and we can give you more specific help.

Unless of course you're checking the box on some form that requires a 'capability study'...in that case you are safe with either formula. : )
 
Q

qakid

Re: Cpk or Ppk?

Thank you for your help. Attached is data that I am working with. The sample size is 30 parts. Every 10th part was pulled from a 300 piece continuous run, so the parts are in the sequential order of which they were produced. There are 13 "X" dimensions on this part. They are all made with the same tool. I am looking at them individually and cumulatively. For my cumulative calculations, each part is a subgroup, so 13 measurements will make up a subgroup. I was looking at the Ppk for each of the 13 locations, and the Cpk for the cumulative capability. I appreciate any help/feedback. Thank you.
 

Attachments

  • Raw Data.xls
    30.5 KB · Views: 223

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Cpk or Ppk?

Attached is data that I am working with. The sample size is 30 parts. Every 10th part was pulled from a 300 piece continuous run, so the parts are in the sequential order of which they were produced. There are 13 "X" dimensions on this part. They are all made with the same tool. I am looking at them individually and cumulatively. For my cumulative calculations, each part is a subgroup, so 13 measurements will make up a subgroup. I was looking at the Ppk for each of the 13 locations, and the Cpk for the cumulative capability.

What is the tolerance? Are these 13 independent measurement on the part (similar features on the same part), or 13 measurements of the same characteristic (such as several measurements of a diameter about the same diameter)?

What is your gage r&r for this dimension? There is very little variation within the reported 4 place resolution. It is not clear if any of the variation is from the process - and that it is all from measurement. Something to consider, unless the tolerance is much larger.

So far, so good - thank you for supplying the data!
 
Q

qakid

Re: Cpk or Ppk?

The Gage R&R %Tolerance is 7.63%. The dimension tolerance is 0.563 +/-.001. This feature is stamped into the part in 13 different locations using the same form tool.
 
Q

qakid

Re: Cpk or Ppk?

The Gage R&R %Tolerance is 7.63%. The dimension tolerance is 0.563 +/-.001. This feature is stamped into the part in 13 different locations using the same form tool.
 
Top Bottom