Subscribe Bookmark RSS Feed

How can I calculate capability index PPK with a USL set at an asymptote?

craigking

Community Member

Joined:

Aug 14, 2014

I have a situation were I need to calculate the PPK for data that is not normally distributed and has an asymptote for the upper specification limit. How can I calculate capability index PPK?

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

Usually we assume a normal distribution when we calculate Ppk; however, when working with skewed data, this assumption will result in a Ppk metric that is often obviously wrong (as you have undoubtedly discovered.)

In JMP there are at least two ways to deal with this (I'm using JMP11)

The "simple" method:

Use Analyze/Distribution. Select your column and the distribution model you would like to use from the dropdown menue in the dialog. After clicking Run, go to the red triangle menu (RTM) and run the capability analysis, and supply at least one spec limit.

The "easy" method:

Go into the column properties of your metric column. Add a distribution property, and a spec limit property. Then just run the distribution (Analyze/Distribution, pick column, click Run.)

In both cases above, I assumed you knew with distribution model.  If you don't know, run a distribution on your metric column, then use the RTM and pick Continuous Fit, and and select "All."  Then pick either model at the top of the list, or the one that makes most sense given your knowledge of the process metric.

2 REPLIES
David_Burnham

Super User

Joined:

Jul 13, 2011

I think there are 3 questions here (1) handling one-sided spec limits (2) handling non-normal data (3) PPK versus CPK.

(1) JMP should handle this automatically.  You just leave the upper spec empty.

(2) By default JMP will use assume a normal distribution.  But from the Distribution platform you can fit say a Log Normal distribution and then Process Capability results will be reported for this distribution.

3) Mathematically CPK and PPK look the same - they differ in what the standard deviation represents.  JMP provides support for PPK in that you can change the label to read PPK instead of CPK - this is done under File>Preferences>Platforms>Distribution: PPK Capability Labeling

Dave

-Dave
Solution

Usually we assume a normal distribution when we calculate Ppk; however, when working with skewed data, this assumption will result in a Ppk metric that is often obviously wrong (as you have undoubtedly discovered.)

In JMP there are at least two ways to deal with this (I'm using JMP11)

The "simple" method:

Use Analyze/Distribution. Select your column and the distribution model you would like to use from the dropdown menue in the dialog. After clicking Run, go to the red triangle menu (RTM) and run the capability analysis, and supply at least one spec limit.

The "easy" method:

Go into the column properties of your metric column. Add a distribution property, and a spec limit property. Then just run the distribution (Analyze/Distribution, pick column, click Run.)

In both cases above, I assumed you knew with distribution model.  If you don't know, run a distribution on your metric column, then use the RTM and pick Continuous Fit, and and select "All."  Then pick either model at the top of the list, or the one that makes most sense given your knowledge of the process metric.