Estimating Sigma for Capability Metrics: Industry Standards?
- This topic has 1 reply, 1 voice, and was last updated 2 years, 6 months ago by Iain Hastings.
- November 26, 2017 at 5:23 pm #55878
RichardParticipant@IronFelix717 Include @IronFelix717 in your post and this person will
be notified via email.
I am back after a year, with another question regarding SPC.
I was recently surprised to learn that my traditional CPK formula was improper. I had simply thought the formulae for the metrics used the sample standard deviation. Oh how terribly misinformed I was….. Misinformed. That seems to be a common theme here in the world of SPC. I’ve never seen a topic so misconstrued and incorrectly taught in my life. Someday the world will realize you can’t try to use statistics if you don’t know anything about them… I digress.
I was rudely awaken with the 6 different estimators for the POPULATION SIGMA, which is utilized in the CP/CPK formulae depending on whether we subgroup or not.
Thus, when subgroup size = 1, sigma is estimated with either:
When subgroup >1, then estimators are based on either…
Some of these, of course, are divided by hartley’s constants…
My question is…
Which is correct? I work in Automotive, so AIAG is preferred. However, i don’t care to open up the AIAG manual at the time, and i’m also curious to hear professional’s opinions on what industries require what. I also understand, the customer may have a unique preference. Whats your experience and which are less common estimators?
-Rich0November 27, 2017 at 7:19 am #202012
The AIAG calculations are based on a “Significant production run” as detailed in the PPAP manual. A significant production run is 300 parts manufactured under production conditions.
The required calculation is for Cpk using the standard deviation estimated using Rbar – this assumes a stable process. The PPAP manual does allow an alternative for an unstable process (with identifiable trends or patterns!) in which case Ppk should be used – ie using the sample standard deviation.
The idea of calculating indices on an unstable process is somewhat contentious – Douglas Montgomery would liken it to “statistical terrorism”. The key point in the AIAG manual is the predictab;e trend or pattern.
Note that although this is what is contained in the AIAG PPAP manual, Customer-Specific Requirements may differ. For example, I believe that Ford requires using the sample standard deviation for calculation.0
You must be logged in to reply to this topic.