Subscribe Bookmark RSS Feed

Cpk with non-normal data

mindjob

Community Trekker

Joined:

Jun 23, 2011

I have JMP 8.0 and read in the used manual that there is a function called "Spec Limits", which will tell me what kind of data I have, and then I can use it to calculate Cpk. Is this correct and how does it work?
4 REPLIES
louv

Staff

Joined:

Jun 23, 2011

If you use the distribution platform and get a distribution of your response then go under the red triangle and utilize >continuous fit >all you will then get an assessment of the fit of your data. Once you do that you can go under the red triangle and choose capability analysis where you would dial in the type of distribution, the lower and upper specification limits and the target value. Once you do this you can compare the Cpk of your data versus the Cpk if you had assumed it to be normally distributed.
outlier

Community Trekker

Joined:

Jun 23, 2011

Hi Lou,

Can yo confirm where I select the distribution..? I try with the check box or via the drop down menu but neither seems to impact the calculated Cpk numbers. I can see the fits for each distribution on the histogram (Weibull, normal, log etc.), but Cpk number is always the same.
louv

Staff

Joined:

Jun 23, 2011

Open JMP Sample Data File = Big Class
Choose Analyze Response = "Weight"; Click OK
Red Triangle on Response; Scroll down to Continuous Fit> "All"
Scroll Down to output; Compare the fit (Log Normal is best)
Go To Red Triangle> Capability Analysis
Add Specification; LSL = 50, USL = 150
In Pull down window on dialog box choose "LogNormal"
Cpk calculated will us Log Normal distribution to calculate

Message was edited by: Lou V
outlier

Community Trekker

Joined:

Jun 23, 2011

Thanks, that worked.!