Ppk, Cpk, or something else?

[gnojham]

Hi all,
looking for some info on how to approach a situation at work.
we have a system that produces a widget.
we want to measure a certain characteristic of the widget and show that the process is capable of producing widgets within spec before we launch the product (making the widgets for sale)
currently, we are using Cpk, but I'm not so sure it is correct, and if it is correct, we aren't calculating it correctly. apparently for Cpk you need control charts and subgroups and such, which we don't have because we aren't making these widgets yet.

Is Ppk the better/correct option, or is there maybe some other test?

I think we made let's say about 5,000 widgets, and about 60 samples were taken, and that's the data we have to work with.

Looking forward to hearing what you all think.

Also, if this is in the wrong spot, please move?
Thanks

 

[gnojham]

Hi all,
I was just reading through the guidelines a bit and came across the part about not using all lowercase. Sorry about that. It's a bad habit of mine.

Also, to further comply with the guidelines:
When I said 'Apparently you need control charts...', that meant that I looked at various resources on the Web and also checked some things in my old 6 sigma literature.
Thanks for any input on this.

 

[Miner]

Welcome @gnojham . I see from your post that at least this particular thread is quality related. While this forum is statistics related, you will find few members that are fluent in quality topics. You will find more responses at the two quality forums below:

• Elsmar Cove
• Quality Forum Online

To answer your question, the difference between Cp/Cpk and Pp/Ppk is the element of time. Cp/Cpk is based on short term process variation (within subgroups), while Pp/Ppk is based on long term variation (within plus between subgroup variation. The intent behind all of these metrics is that they are based upon actual data from a significant production run (quantity plus time). 5000 certainly qualifies as a significant product run from a quantity perspective whereas an 8-hour shift is expected for the time element.

• Was this a pilot run? How closely would it resemble actual production (i.e., what normal sources of variation were not present in this run that would be present in production)?
• Were the 60 samples taken in time order sequence?
• Were they sub-grouped, taken individually over time, taken randomly, or all at once?