CPK is not true if data is not normal is PPK true?
Six Sigma – iSixSigma › Forums › Old Forums › General › CPK is not true if data is not normal is PPK true?
- This topic has 7 replies, 7 voices, and was last updated 14 years, 7 months ago by
Van Kim Ban.
-
AuthorPosts
-
November 20, 2007 at 1:57 pm #48724
Scott JohnsonMember@Scott-JohnsonInclude @Scott-Johnson in your post and this person will
be notified via email.My question is related to data that is tested to not be normal using the Anderson Darling test. For Capablity of a process index (CPK) is not relevent if data is not normally distributed. Is Performance of the process index (PPK) a relevent number or mathmatically correct?
0November 20, 2007 at 2:18 pm #165168
Search ForumMember@Search-ForumInclude @Search-Forum in your post and this person will
be notified via email.November 20, 2007 at 3:43 pm #165173Your first statement is incorrect. Cp-Cpk-Pp-Ppk-Z-Sigma all require an approximately normal distribution to be accurate. The difference between the Cp and Pp is simply the type of std dev you are using in your calcuation.
0November 20, 2007 at 6:29 pm #165184I like the “posted by” name, but I don’t think many get it.
0November 20, 2007 at 6:39 pm #165185
freshmanParticipant@freshmanInclude @freshman in your post and this person will
be notified via email.Hi Pete,
This question is very important for me. Can you give us some example data’s ? Because i want to try this point on Minitab.0November 20, 2007 at 9:09 pm #165189
Van Kim BanMember@Van-Kim-BanInclude @Van-Kim-Ban in your post and this person will
be notified via email.long-term or short-term
0November 20, 2007 at 9:54 pm #165190
RealityParticipant@RealityInclude @Reality in your post and this person will
be notified via email.In reality, you will probably never run accross a perfect bell curve quantity of data. Therefore, in most all applications, you will be applying bell curves to data that may be near, but does not have a perfect distribution of data to fit a bell curve. With this in mind, I will direct you to Dr. Donald Wheeler’s book “Introduction to Statistical Process Control” where he points out that slight normallity diviations has very little affect on the predictability of the statistics…even to the extent that the curve may be mis-shaped as badly as a right triangle curve. Therefore, “you” need to decide how critical your application is and whether you want to apply a bell curve to it. You may want to use the Chi-square test or the Anderson-Darling “p” to help you decide. Of course, any good stats software package today will also calculate the proper CpK for you if you tell it to use curve fitting (a non-normal curve) to do the calculations.
0November 21, 2007 at 8:34 am #165202
Van Kim BanMember@Van-Kim-BanInclude @Van-Kim-Ban in your post and this person will
be notified via email.Thank You.
I wish to purchase that book,kindly send me some details
Best regards0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.