Why JMP Cpk (Ppk) is not consistent with %non-conformance?

[#1]

Hi.

I calculated the nonparametric Cpk as below.

And JMP shows Cpk as Ppk in the report 

(that is, so called Cpk is shown as Ppk in JMP)

However, Cpk 0.667 is equal to;

1) % conformance: 68.27%

Cpk 1 is equal to;

1) % conformance: 95.45%

Thus I can think of this 0.791 Cpk as about 80% of conformance.

Then this is the point where I am confused to the result of the lowest part of the report.

Nonconformance Expected Overall %= 0%?

How can I accept this?

Please let me know how to interpret this.

Thank you.

[#2]

Cpk is estimated by using short-term sigma (using estimated sigma: moving average/average range/averag standard deviation). Ppk is estimated by using long-term sigma (estimated sigma: overall dispersion, standard deviation that you see in distribution).

I wont dwell to explain how suddenly Cpk becoming Ppk in recent years...too many past discussion on this

But, I will only refer to Ppk for my process control. Simple to explain.

[#3]

Observed non-conformance refers to your existing data. It is 0 because your data contained no nonconformance. The expected non-conformance at Ppk = 0.791 is 0.5796%. Graphically, it makes sense.

Ppk measure the distance between your spec limits and your observed data. The narrower the gap between the data distribution and your spec limit, you will have lower Ppk.

[#4]

Thanks for help.

However I did not input the historical sigma and there is nothing can be conceieved as long-term sigma.

I don't want to see the Ppk now and literaly I cannot get the PPK with this current, non-subgrouped, few data.

And there were a lot of references that PPK is not the distant term to the CPK, they share the interpretation and meaning. (if same value, same expected extent of nonconformance)

And CPK of 0.67 definetely indicates about 30% nonconformance. (

>This does not consistent with the Reported Expected overall % nonconformance

>There are Consistency among; Figure, Off-specification proportion, Expected overall % nonconformance,... etc. But PPK value itself.

[#5]

Hi Mike,

That long-term sigma refers to standard deviation of your dataset.

Measure of dispersion (sigma) can refers to moving range, average range, average standard deviation, levey-jennings standard deviation.

Cpk only works if you data is normal (since +/- is equal). But, you data is not normal

[#6]

Thank you for answer...

I am just wonder why JMP shows CPK ...? as long as if it cannot be calculated for the non-normal data..?

[#7]

JMP won't calculate Cpk for nonnormal distributions Quality and Process Methods > Process Capability > Process Capability Platform Options > Individual ... 

Note: Capability indices based on within-subgroup variation and stability indices are not available for processes for which you have specified nonnormal distributions.

More on statistics Quality and Process Methods > Process Capability > Statistical Details for the Process Capability Pl...